WO2019000391A1 - 车辆的控制方法、装置及设备 - Google Patents

车辆的控制方法、装置及设备 Download PDF

Info

Publication number
WO2019000391A1
WO2019000391A1 PCT/CN2017/091095 CN2017091095W WO2019000391A1 WO 2019000391 A1 WO2019000391 A1 WO 2019000391A1 CN 2017091095 W CN2017091095 W CN 2017091095W WO 2019000391 A1 WO2019000391 A1 WO 2019000391A1
Authority
WO
WIPO (PCT)
Prior art keywords
self
driving
lane
vehicle
preset
Prior art date
Application number
PCT/CN2017/091095
Other languages
English (en)
French (fr)
Inventor
许松岑
郑荣福
王靓伟
庄雨铮
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to PCT/CN2017/091095 priority Critical patent/WO2019000391A1/zh
Priority to EP17915345.7A priority patent/EP3647140B1/en
Priority to CN201780092791.7A priority patent/CN110809542B/zh
Publication of WO2019000391A1 publication Critical patent/WO2019000391A1/zh
Priority to US16/730,444 priority patent/US20200139989A1/en

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles
    • B60W60/001—Planning or execution of driving tasks
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W10/00—Conjoint control of vehicle sub-units of different type or different function
    • B60W10/04—Conjoint control of vehicle sub-units of different type or different function including control of propulsion units
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W10/00—Conjoint control of vehicle sub-units of different type or different function
    • B60W10/04—Conjoint control of vehicle sub-units of different type or different function including control of propulsion units
    • B60W10/06—Conjoint control of vehicle sub-units of different type or different function including control of propulsion units including control of combustion engines
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W10/00—Conjoint control of vehicle sub-units of different type or different function
    • B60W10/18—Conjoint control of vehicle sub-units of different type or different function including control of braking systems
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W10/00—Conjoint control of vehicle sub-units of different type or different function
    • B60W10/20—Conjoint control of vehicle sub-units of different type or different function including control of steering systems
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
    • B60W30/18—Propelling the vehicle
    • B60W30/18009—Propelling the vehicle related to particular drive situations
    • B60W30/18163—Lane change; Overtaking manoeuvres
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/0088—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
    • B60W2050/0001—Details of the control system
    • B60W2050/0019—Control system elements or transfer functions
    • B60W2050/0028—Mathematical models, e.g. for simulation
    • B60W2050/0031—Mathematical model of the vehicle
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
    • B60W2050/0062—Adapting control system settings
    • B60W2050/0075—Automatic parameter input, automatic initialising or calibrating means
    • B60W2050/0083—Setting, resetting, calibration
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W2552/00—Input parameters relating to infrastructure
    • B60W2552/53—Road markings, e.g. lane marker or crosswalk
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W2710/00—Output or target parameters relating to a particular sub-units
    • B60W2710/06—Combustion engines, Gas turbines
    • B60W2710/0605—Throttle position
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W2720/00—Output or target parameters relating to overall vehicle dynamics
    • B60W2720/10—Longitudinal speed

Definitions

  • the present invention relates to the field of communications technologies, and in particular, to a method, device, and device for controlling a vehicle.
  • the automatic driving system is an intelligent system integrating automation, architecture, artificial intelligence, visual computing and many other technologies.
  • the control decision system in the automatic driving system can determine the strategy of automatic driving according to the information analyzed by the sensing system, and then output corresponding control commands to realize the purpose of planning the vehicle driving route and controlling the automatic driving of the vehicle.
  • the control decision system needs to take into account the safety and comfort of the car when determining the automatic driving strategy, and needs to ensure the safety of the driving process while reaching the destination as soon as possible. Therefore, the automatic driving system has strict requirements for the safety and reliability of the control decision system to control the automatic driving of the vehicle.
  • the control decision system in order to ensure the safety and reliability of the vehicle when driving automatically, the control decision system is usually based on various parameters of the self-driving vehicle while driving, and uses a learning algorithm to calculate and output a control command of the self-driving vehicle, such as a steering wheel angle. , throttle size, braking force, etc., control the self-driving vehicle through the output control command.
  • a control command of the self-driving vehicle such as a steering wheel angle. , throttle size, braking force, etc.
  • the self-driving vehicle often has a situation that does not conform to the natural driving behavior. For example, the self-driving vehicle frequently swings left and right, and even causes a safety accident, so it will be automatically Driving a vehicle poses a safety hazard.
  • the embodiment of the invention provides a method, a device and a device for controlling a vehicle, which can solve the problem that a safety hazard is brought to a vehicle that is automatically driven when the vehicle is automatically driven by a control command in the prior art.
  • an embodiment of the present invention provides a method for controlling a vehicle, including:
  • the coarse-grained control command is used to control the driving mode of the self-driving vehicle, and the control command decision model is based on the self-driving vehicle
  • the training driving state information in the training state and the training driving environment information training of the self-driving vehicle in the training state are obtained;
  • the second driving environment information includes driving environment information corresponding to the coarse-grained control instruction when the self-driving vehicle is in the automatic driving state
  • the fine-grained control instruction corresponding to the coarse-grained control instruction is determined according to the lane information and the driving state information of the self-driving vehicle, and the fine-grained control instruction is used to control the driving parameter of the self-driving vehicle, the lane information
  • the coarse-grained control command is first calculated by using the driving state information, the first driving environment information, and the control command decision model, and then determining whether the coarse-grained control command can be executed based on the second driving environment information, that is, determining the control Before driving the driving parameters of the vehicle, the coarse-grained control command is calculated to calculate the driving mode of the self-driving vehicle.
  • the coarse-grained control command is converted into the fine-grained control command, thereby avoiding Outputting an unnecessary or erroneous control command during automatic vehicle control to prevent the self-driving vehicle from appearing to be inconsistent with the natural driving behavior or transmitting a safety accident; meanwhile, in the embodiment of the present invention, after determining that the coarse-grained control command can be executed, the vehicle is automatically driven.
  • the driving state information and the lane information corresponding to the coarse-grained control command convert the coarse-grained control command into a fine-grained control command, that is, fully consider the self-driving state of the self-driving vehicle and the lane information to determine a fine-grained control suitable for controlling the self-driving vehicle.
  • instruction Accuracy of high-fine-grain control commands and accuracy of autonomous vehicle control, avoiding auto-driving vehicles that do not conform to natural driving behavior or sending safety incidents, reducing safety hazards of self-driving vehicles and further improving the safety of self-driving vehicles And comfort.
  • the coarse-grained control instruction comprises a straight line
  • the lane information includes a preset desired speed of the lane in which the self-driving vehicle travels straight, and the driving state information of the self-driving vehicle includes the current speed of the self-driving vehicle;
  • Determining the fine-grained control instruction corresponding to the coarse-grained control instruction according to the lane information and the driving state information of the self-driving vehicle including:
  • a fine-grained control command is determined based on the preset desired speed and the current speed.
  • the steering wheel angle of the self-driving vehicle does not need to be controlled, and only the traveling speed of the self-driving vehicle can be controlled, so the vehicle is traveling straight according to the current speed of the self-driving vehicle and the self-driving vehicle.
  • the preset desired speed of the lane is to determine the fine-grained control command for the self-driving vehicle, improve the accuracy of the fine-grained control command and the accuracy of the self-driving vehicle control, and avoid the unnatural driving behavior of the self-driving vehicle.
  • the fine-grained control instruction comprises a throttle size of the self-driving vehicle
  • the throttle size is equal to zero
  • the throttle size of the self-driving vehicle is calculated based on the first preset formula, wherein the first preset formula includes:
  • Throttle size preset throttle control coefficient ⁇ (preset desired speed - current speed + preset value).
  • the throttle size is determined to be zero at this time, so as to reduce the running speed of the self-driving vehicle as soon as possible; the current speed is not greater than the pre-pre-
  • the desired speed is set, it indicates that the driving speed of the self-driving vehicle is normal.
  • the throttle size is determined based on the difference between the preset desired speed and the current speed and the preset throttle control coefficient, so as to achieve precise control of the throttle and ensure automatic driving. The safety of the vehicle.
  • the preset desired speed is not greater than the maximum speed limit of the lane in which the self-driving vehicle is traveling straight.
  • Fine-grained control commands include the braking force of an autonomously driven vehicle
  • the braking force is equal to zero
  • the braking force of the self-driving vehicle is calculated based on the second preset formula, wherein the second preset formula includes:
  • Brake force preset brake control coefficient ⁇ (current speed - preset desired speed).
  • the driving speed of the self-driving vehicle when the current speed is less than the preset desired speed, the driving speed of the self-driving vehicle is normal. At this time, the braking may not be performed, so the braking force is adjusted to zero; when the current speed is not less than the preset desired speed, the automatic is described. The driving speed of the driving vehicle is too large. At this time, the speed of the self-driving vehicle needs to be controlled by the brake, so the braking force is determined based on the difference between the preset desired speed and the current speed and the preset braking control coefficient to achieve the braking. Precise control ensures the safety of self-driving vehicles.
  • the preset desired speed is not greater than the maximum speed limit of the lane in which the self-driving vehicle is traveling straight.
  • the coarse-grained control instruction includes a lane change direction
  • Determining the fine-grained control instruction corresponding to the coarse-grained control instruction according to the lane information and the driving state information of the self-driving vehicle including:
  • a fine-grained control command of the self-driving vehicle is determined based on the lane change travel route and the lane information.
  • the automatic driving vehicle when the coarse-grained control command includes the direction change direction, the automatic driving vehicle is controlled to change lanes to the lane changing direction side.
  • the lane changing driving path of the self-driving vehicle is first simulated, that is, the automatic driving vehicle is actually changed.
  • the lane change path is determined, and then the fine-grained control command of the self-driving vehicle during the lane change is determined based on the predetermined lane change route and lane information, thereby improving the accuracy of the fine-grained control command and the automatic driving vehicle control. Accuracy, which in turn improves the safety of self-driving vehicles when changing lanes.
  • the fine-grain control command includes a steering wheel angle
  • the lane information includes a lane width corresponding to the lane of the lane change travel route and a lane centerline corresponding to the lane of the lane change travel route;
  • Determining fine-grained control commands for the self-driving vehicle based on the lane change travel route and lane information including:
  • the steering wheel angle is calculated based on the third preset formula, the steering angle, the target lane width, and the target distance, wherein the third preset formula includes:
  • Steering wheel angle [steering angle - target distance / (target lane width ⁇ first corner factor)] ⁇ second corner coefficient.
  • the lane change travel path of the self-driving vehicle is a curve
  • the steering wheel angle of the self-driving vehicle needs to be controlled, it is possible to determine the need for the self-driving vehicle to be discrete when changing lanes based on the pre-simulated lane change travel route. Steering position and steering angle, and then determining the angle of rotation of the steering wheel corresponding to each steering angle in combination with the information of each steering angle corresponding to the lane in the lane changing driving path, that is, determining the angle at which the steering wheel needs to be rotated when the autonomous vehicle is changing lanes, Thereby achieving precise control of the steering wheel and ensuring the safety of driving the vehicle automatically.
  • the preset desired speed is not greater than the maximum speed limit of the lane in which the self-driving vehicle is traveling straight.
  • the value of the steering wheel angle can be greater than or equal to -1 and less than or equal to 1.
  • the fine-grain control command further includes a throttle size of the self-driving vehicle
  • the lane information includes a preset desired speed of the lane corresponding to the lane change driving path;
  • the method further includes:
  • the throttle size is equal to zero
  • the throttle size of the auto-driving vehicle at the corresponding position of the steering angle is calculated based on the fourth preset formula, wherein the fourth preset formula includes:
  • Throttle size preset throttle control coefficient ⁇ (target preset desired speed - current speed + preset value).
  • the traveling speed can also be controlled.
  • the throttle size is determined to be zero, so as to reduce the driving speed of the self-driving vehicle as soon as possible; when the current speed is not greater than the target preset desired speed, the target preset desired speed and the current speed are determined based on the target.
  • the difference between the difference and the preset throttle control coefficient determines the throttle size to achieve precise control of the throttle while ensuring the safety of the self-driving vehicle.
  • the target preset desired speed is not greater than the highest speed limit of the lane corresponding to the lane change travel path.
  • the fine-grain control command further includes braking force of the self-driving vehicle
  • the lane information further includes a preset desired speed of the lane corresponding to the lane change driving path;
  • the method further includes:
  • the braking force is equal to zero
  • the braking force of the self-driving vehicle at the corresponding position of the steering angle is calculated based on the fifth preset formula, wherein the fifth preset formula includes:
  • Brake force preset brake control coefficient ⁇ (current speed - target preset desired speed).
  • the traveling speed can also be controlled.
  • the brake may not be performed, so the braking force is adjusted to zero;
  • the target preset is based on the target
  • the difference between the desired speed and the current speed and the preset brake control coefficient determine the braking force to achieve precise control of the brakes and ensure the safety of the self-driving vehicle.
  • the target preset desired speed is not greater than the highest speed limit of the lane corresponding to the lane change travel path.
  • the throttle size can range from 0 to greater than 1 and less than or equal to 1.
  • the braking force can be in the range of 0 or more and less than or equal to 1.
  • the preset value can be 1.
  • the first driving environment information includes lane information of a lane in which the self-driving vehicle travels, and the vehicle is automatically driven within a preset distance of the vehicle.
  • Information at least one of the information of the road surface within the preset distance of the self-driving vehicle;
  • the second driving environment information includes lane information corresponding to the coarse-grained control instruction in the driving lane of the self-driving vehicle, and the vehicle information corresponding to the coarse-grained control instruction within the preset distance of the self-driving vehicle, and the preset distance and coarse granularity of the self-driving vehicle At least one of the road surface information corresponding to the control command.
  • the second driving environment information includes a first vehicle distance between the target vehicle and the self-driving vehicle, and the target vehicle indicates automatic driving a vehicle corresponding to the coarse-grained control command within the preset distance of the vehicle;
  • Determining whether to execute the coarse-grained control instruction according to the second driving environment information including:
  • the granularity control instruction determines whether the first vehicle distance between the target vehicle and the self-driving vehicle reaches a safe distance.
  • the coarse-grained control instruction includes a lane change direction
  • the method further includes:
  • the second driving environment information includes a second vehicle distance when the target vehicle and the self-driving vehicle are traveling on the lane changing driving path, and the target vehicle indicates the vehicle corresponding to the coarse grain control instruction within the preset distance of the autonomous driving vehicle;
  • Determining whether to execute the coarse-grained control instruction according to the second driving environment information including:
  • the lane-changing driving path of the self-driving vehicle may be simulated first, and then based on whether the target vehicle and the second vehicle distance when the self-driving vehicle is traveling on the lane-changing driving route
  • the coarse-grained control command is executed to judge, thereby improving the accuracy of the judgment and the safety of the self-driving vehicle in automatic driving.
  • the safety distance may be: the self-driving vehicle runs at the current speed of the self-driving vehicle, and the target vehicle runs at the current speed of the target vehicle, and the self-driving vehicle and the target vehicle are made within a preset time period. The distance between no contact.
  • calculating a coarse granularity of the self-driving vehicle based on the driving state information, the first driving environment information, and the control instruction decision model Before the control instruction it also includes:
  • training parameters including training driving state information and training driving environment information
  • the corresponding model parameter is determined as the final model parameter of the control instruction decision model.
  • the loss function Loss1 includes:
  • Loss1
  • v represents the current speed of the self-driving vehicle
  • ⁇ represents the angle between the current driving direction of the self-driving vehicle and the lane where the self-driving vehicle is located
  • Q represents the coarse-grained training control command and the training parameter corresponding to the preset coarse-grained training control command. The degree of matching between.
  • the method before updating the model parameter, the method further includes:
  • Update model parameters including:
  • the updated model parameters are calculated based on the update gradient, the preset update coefficient, and the model parameters before the update.
  • the method before calculating the update gradient of the model parameter, the method further includes:
  • an embodiment of the present invention provides a control device for a vehicle, including:
  • An acquiring unit configured to acquire driving state information of the self-driving vehicle and first driving environment information of the self-driving vehicle when the autonomous driving vehicle is in an automatic driving state;
  • a calculation unit configured to calculate a coarse-grained control instruction of the self-driving vehicle based on the driving state information, the first driving environment information, and the control instruction decision model, and the coarse-grained control instruction is used to control the driving mode of the self-driving vehicle, and the control instruction decision model is The training driving state information based on the self-driving vehicle in the training state and the training driving environment information when the self-driving vehicle is in the training state are trained;
  • a determining unit configured to determine, according to the second driving environment information, whether to execute a coarse-grained control instruction, where the second driving environment information includes driving environment information corresponding to the coarse-grained control instruction when the self-driving vehicle is in an automatic driving state;
  • a determining unit configured to determine a fine-grained control instruction corresponding to the coarse-grained control instruction according to the lane information and the driving state information of the self-driving vehicle when determining to execute the coarse-grained control instruction, where the fine-grained control instruction is used to control the self-driving vehicle
  • the driving parameter, the lane information includes information corresponding to the lane of the coarse-grained control instruction in the road on which the self-driving vehicle is traveling;
  • An output unit for outputting fine-grained control instructions.
  • the control device first calculates the coarse-grained control command by using the driving state information, the first driving environment information, and the control command decision model, and then determines whether the coarse-grained control command can be executed based on the second driving environment information, that is, Before determining the driving parameters of the self-driving vehicle, the coarse-grained control command is calculated to calculate the driving mode of the self-driving vehicle, and when it is determined that the coarse-grained control command can be executed, the coarse-grained control command is converted into the fine-grained control command, thereby It is possible to prevent unnecessary or erroneous control commands from being output during the automatic control of the vehicle, and to prevent the self-driving vehicle from being inconsistent with the natural driving behavior or to transmit a safety accident.
  • the coarse-grained control instruction includes a straight line
  • the lane information includes a preset desired speed of the lane in which the self-driving vehicle travels straight, and the driving state information of the self-driving vehicle includes the current speed of the self-driving vehicle;
  • the determining unit is specifically configured to determine the fine-grained control instruction based on the preset desired speed and the current speed.
  • the control device when the coarse-grained control command includes straight traveling, the control device does not need to control the steering wheel angle of the self-driving vehicle, and can control only the traveling speed of the self-driving vehicle, so according to the current speed of the self-driving vehicle and the self-driving vehicle Predetermined speed in the straight lane to determine fine-grained control commands for self-driving vehicles, improving the accuracy of fine-grained control commands and the accuracy of autonomous vehicle control, Avoid auto-driving vehicles that appear to swing side to side, etc. that do not conform to natural driving behavior.
  • the fine-grained control instruction comprises a throttle size of the self-driving vehicle
  • the determining unit is specifically used for:
  • the throttle size is equal to zero
  • the throttle size of the self-driving vehicle is calculated based on the first preset formula, wherein the first preset formula includes:
  • Throttle size preset throttle control coefficient ⁇ (preset desired speed - current speed + preset value).
  • the control device determines the throttle size to be zero at this time, so as to reduce the traveling speed of the self-driving vehicle as soon as possible; the current speed is not When the preset speed is greater than the preset desired speed, the driving speed of the self-driving vehicle is normal. At this time, the control device determines the throttle size based on the difference between the preset desired speed and the current speed and the preset throttle control coefficient, so as to achieve precise control of the throttle. To ensure the safety of driving autonomous vehicles.
  • the preset desired speed is not greater than the maximum speed limit of the lane in which the self-driving vehicle is traveling straight.
  • the fine-grained control command includes braking force of the self-driving vehicle
  • the determining unit is specifically used for:
  • the braking force is equal to zero
  • the braking force of the self-driving vehicle is calculated based on the second preset formula, wherein the second preset formula includes:
  • Brake force preset brake control coefficient ⁇ (current speed - preset desired speed).
  • the control device may not perform braking, so the braking force is adjusted to zero; when the current speed is not less than the preset desired speed, It indicates that the driving speed of the self-driving vehicle is too large.
  • the control device needs to control the speed of the self-driving vehicle through the brake, so the braking force is determined based on the difference between the preset desired speed and the current speed and the preset braking control coefficient. Achieve precise control of the brakes to ensure the safety of the self-driving vehicle.
  • the preset desired speed is not greater than the maximum speed limit of the lane in which the self-driving vehicle is traveling straight.
  • the coarse-grain control command includes a lane change direction
  • the determining unit is specifically used for:
  • a fine-grained control command of the self-driving vehicle is determined based on the lane change travel route and the lane information.
  • the control device controls the automatic driving vehicle to perform the lane change to the lane changing direction side.
  • the lane changing driving path of the autonomous driving vehicle is first simulated, that is, the actual driving vehicle is actually driven.
  • the lane information determines the fine-grained control command of the self-driving vehicle during the lane change, thereby improving the accuracy of the fine-grained control command and the accuracy of the self-driving vehicle control, thereby improving the safety of the self-driving vehicle when changing lanes.
  • the fine-grain control command includes a steering wheel angle
  • the lane information includes a lane width corresponding to the lane of the lane change travel route and a lane centerline corresponding to the lane of the lane change travel route;
  • the determining unit is specifically used for:
  • the steering wheel angle is calculated based on the third preset formula, the steering angle, the target lane width, and the target distance, wherein the third preset formula includes:
  • Steering wheel angle [steering angle - target distance / (target lane width ⁇ first corner factor)] ⁇ second corner factor.
  • the control device since the lane change travel path of the self-driving vehicle is a curve, the control device needs to control the steering wheel angle of the self-driving vehicle. Therefore, based on the pre-simulated lane change travel route, it can be determined that the self-driving vehicle needs to be changed when changing lanes. Steering position and steering angle, and then determining the angle of rotation of the steering wheel corresponding to each steering angle in combination with the information of each steering angle corresponding to the lane in the lane changing driving path, that is, determining the angle at which the steering wheel needs to be rotated when the autonomous vehicle is changing lanes, Thereby achieving precise control of the steering wheel and ensuring the safety of driving the vehicle automatically.
  • the preset desired speed is not greater than the maximum speed limit of the lane in which the self-driving vehicle is traveling straight.
  • the value of the steering wheel angle can be greater than or equal to -1 and less than or equal to 1.
  • the fine-grain control command further includes a throttle size of the self-driving vehicle
  • the lane information further includes a preset desired speed of the lane corresponding to the lane change driving path;
  • the determining unit is specifically used for:
  • the throttle size is equal to zero
  • the throttle size of the auto-driving vehicle at the corresponding position of the steering angle is calculated based on the fourth preset formula, wherein the fourth preset formula includes:
  • Throttle size preset throttle control coefficient ⁇ (target preset desired speed - current speed + preset value).
  • the control device can also control the traveling speed when the self-driving vehicle changes lanes.
  • the throttle size is determined to be zero, so as to reduce the automatic as soon as possible Driving speed of the driving vehicle; when the current speed is not greater than the target preset desired speed, the throttle size is determined based on the difference between the target preset desired speed and the current speed and the preset throttle control coefficient to achieve precise control of the throttle while Ensure the safety of self-driving vehicles.
  • the target preset desired speed is not greater than the highest speed limit of the lane corresponding to the lane change travel path.
  • the fine-grain control command further includes braking force of the self-driving vehicle
  • the lane information further includes a preset desired speed of the lane corresponding to the lane change driving path;
  • the determining unit is specifically used for:
  • the braking force is equal to zero
  • the braking force of the self-driving vehicle at the corresponding position of the steering angle is calculated based on the fifth preset formula, wherein the fifth preset formula includes:
  • Brake force preset brake control coefficient ⁇ (current speed - target preset desired speed).
  • the control device can also control the traveling speed when the self-driving vehicle changes lanes.
  • the brake may not be performed, so the braking force is adjusted to zero;
  • the target preset is based on the target
  • the difference between the desired speed and the current speed and the preset brake control coefficient determine the braking force to achieve precise control of the brakes and ensure the safety of the self-driving vehicle.
  • the target preset desired speed is not greater than the highest speed limit of the lane corresponding to the lane change travel path.
  • the throttle size can range from 0 to greater than 1 and less than or equal to 1.
  • the braking force can be in the range of 0 or more and less than or equal to 1.
  • the preset value can be 1.
  • the first driving environment information includes lane information of a lane in which the self-driving vehicle travels, and the vehicle is automatically driven within a preset distance of the vehicle.
  • Information at least one of the information of the road surface within the preset distance of the self-driving vehicle;
  • the second driving environment information includes lane information corresponding to the coarse-grained control instruction in the driving lane of the self-driving vehicle, and the vehicle information corresponding to the coarse-grained control instruction within the preset distance of the self-driving vehicle, and the preset distance and coarse granularity of the self-driving vehicle At least one of the road surface information corresponding to the control command.
  • the second driving environment information includes a first vehicle distance between the target vehicle and the self-driving vehicle, and the target vehicle indicates automatic driving a vehicle corresponding to the coarse-grained control command within the preset distance of the vehicle;
  • the judgment unit is specifically used for:
  • the control device can determine whether the vehicle corresponding to the coarse-grained control command within the preset distance of the self-driving vehicle affects the automatic driving.
  • the safety of the vehicle is determined to determine whether to execute the coarse-grained control command, that is, whether the first vehicle distance between the target vehicle and the self-driving vehicle reaches a safe distance.
  • the coarse-grained control instruction is not executed when the first vehicle distance does not reach the safety distance, and the coarse-grained control instruction is executed when the first vehicle distance reaches the safety distance, which can be effective Improve the safety of self-driving vehicles in automatic driving, avoiding the occurrence of safety accidents caused by the execution of incorrect or unnecessary control commands by the self-driving vehicle.
  • the coarse-grain control command includes a lane change direction
  • the device further includes:
  • An analog unit configured to simulate, according to driving state information of the self-driving vehicle, a lane changing driving path of the self-driving vehicle in a lane changing direction;
  • the second driving environment information includes a second vehicle distance when the target vehicle and the self-driving vehicle are traveling on the lane changing driving path, and the target vehicle indicates the vehicle corresponding to the coarse grain control instruction within the preset distance of the autonomous driving vehicle;
  • the judgment unit is specifically used for:
  • the control device may first simulate the lane change travel path of the self-driving vehicle, and then based on the second vehicle distance when the target vehicle and the self-driving vehicle travel on the lane change travel route Judging whether to execute coarse-grained control commands, thereby improving the accuracy of the judgment and the safety of the self-driving vehicle in automatic driving.
  • the safety distance may be: the self-driving vehicle runs at the current speed of the self-driving vehicle, and the target vehicle runs at the current speed of the target vehicle, and the self-driving vehicle and the target vehicle are made within a preset time period. The distance between no contact.
  • the method further includes:
  • An initialization unit configured to initialize model parameters of the control instruction decision model
  • the obtaining unit is further configured to acquire training parameters, where the training parameters include training driving state information and training driving environment information;
  • the calculating unit is further configured to calculate a coarse-grained training control instruction of the self-driving vehicle according to the model parameter and the training parameter; and, for calculating a value of the loss function according to the coarse-grained training control instruction;
  • the device also includes:
  • An update unit configured to update a model parameter when a value of the loss function does not reach a preset condition
  • the calculating unit is further configured to calculate an updated coarse-grained training control instruction of the self-driving vehicle according to the updated model parameter and the training parameter, and recalculate the value of the loss function until the value of the loss function reaches a preset condition;
  • the determining unit is further configured to determine the corresponding model parameter as the final model parameter of the control instruction decision model when the value of the loss function reaches a preset condition.
  • the loss function Loss1 includes:
  • Loss1
  • v denotes the current speed of the self-driving vehicle
  • ⁇ denotes the angle between the current driving direction of the self-driving vehicle and the lane where the self-driving vehicle is located
  • Q denotes a coarse-grained training control command corresponding to the training parameter Set the degree of matching between coarse-grained training control instructions.
  • the calculating unit is further configured to calculate an update gradient of the model parameter
  • the updating unit is specifically configured to calculate the updated model parameters based on the update gradient, the preset update coefficient, and the model parameters before the update.
  • the determining unit is further configured to determine whether to execute the coarse-grained training control instruction
  • the calculation unit is specifically used to:
  • an embodiment of the present invention provides a control device for a vehicle, including:
  • the memory and the processor are connected by a bus and complete communication with each other;
  • the memory is used to store program code
  • the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory for performing the method as described in the first aspect.
  • an embodiment of the present invention provides a computer readable storage medium, where the computer readable storage medium stores instructions that, when run on a computer, cause the computer to perform the method described in the first aspect.
  • an embodiment of the present invention provides a computer program product comprising instructions that, when run on a computer, cause the computer to perform the method of the first aspect.
  • an embodiment of the present invention provides a computer program that, when run on a computer, causes the computer to perform the method of the first aspect.
  • FIG. 1 is a schematic flowchart of a method for controlling a vehicle according to an embodiment of the present invention
  • FIG. 2 is a schematic flow chart of a method for determining a fine-grained control instruction according to an embodiment of the invention
  • FIG. 3 is a schematic flow chart of still another method for determining fine-grained control instructions according to an embodiment of the invention.
  • FIG. 4 is a schematic block diagram of a control device for a vehicle according to an embodiment of the present invention.
  • FIG. 5 is a schematic block diagram of a control device for a vehicle according to still another embodiment of the present invention.
  • FIG. 6 is a schematic block diagram of a control device of a vehicle according to an embodiment of the present invention.
  • Embodiments of the present invention are applicable to scenarios in which automatic driving control is performed on a vehicle.
  • the control of the self-driving vehicle in the embodiment of the present invention may be a control decision system, and the control decision system may be divided into two parts, namely, a reinforcement learning layer and a motion track security control layer.
  • the reinforcement learning layer can calculate the coarse-grained control command based on the acquired driving-driven vehicle-related information, and then transmit the coarse-grained control command to the motion trajectory security control layer; the motion trajectory security control layer can be safe from the driving The angle determines whether to execute the coarse-grained control instruction.
  • the coarse-grained control instruction is converted into a fine-grained control instruction, that is, the specific driving parameter for controlling the driving of the self-driving vehicle is controlled, and the fine-grained control command is output to The vehicle's control system to achieve automatic driving control of the vehicle.
  • FIG. 1 shows a schematic flow chart of a control method of a vehicle according to an embodiment of the present invention. As shown in FIG. 1, the method includes the following steps 101-105.
  • the driving state information of the self-driving vehicle and the first driving environment information may be acquired in real time, and then the coarse-grained size of the self-driving vehicle is calculated based on the acquired driving state information and the first driving environment information. Control instruction.
  • the first driving environment information may include lane information of a lane in which the vehicle is driven, at least one of information of the vehicle within the preset distance of the vehicle, and information of the road surface within the preset distance of the vehicle.
  • the lane information may include a maximum speed limit of the lane and a width of the lane, etc.
  • the information of the vehicle may include the number of vehicles, the direction of travel of the vehicle, and the distance between the vehicle and the self-driving vehicle, etc.
  • the information of the road surface may include the isolation facility on the road surface. Information and information on obstacles on the road, etc.
  • the driving state information of the self-driving vehicle may include a position of the self-driving vehicle, a speed of the self-driving vehicle, a traveling direction of the self-driving vehicle, and an angle between the self-driving vehicle and the lane in which the vehicle is traveling, and the like.
  • the manner of acquiring the driving state information and the first driving environment information is not limited, and may include various types of sensors, such as a laser radar, an ultrasonic radar, a millimeter wave radar, etc., an in-vehicle camera, and a global positioning system ( Global Positioning System (GPS), maps, on-board diagnostic system (OBD) data for autonomous vehicles, and more.
  • sensors such as a laser radar, an ultrasonic radar, a millimeter wave radar, etc., an in-vehicle camera, and a global positioning system ( Global Positioning System (GPS), maps, on-board diagnostic system (OBD) data for autonomous vehicles, and more.
  • GPS Global Positioning System
  • OBD on-board diagnostic system
  • the coarse-grained control command is used to control the driving mode of the self-driving vehicle
  • the control command decision model is based on the training driving state information of the self-driving vehicle in the training state and the training driving environment information of the self-driving vehicle in the training state.
  • the driving mode may include a straight line, a lane change, a U-turn, a turn, and the like.
  • the coarse-grained control command may include a straight line or a lane change.
  • the coarse-grained control command includes a lane change, the lane change direction may also be included.
  • the control command decision model is pre-trained, which takes training driving state information and training driving environment information as inputs during training.
  • the second driving environment information includes driving environment information corresponding to the coarse-grain control command when the self-driving vehicle is in the automatic driving state.
  • the driving mode of the self-driving vehicle has been determined after the step 102 calculates the coarse-grained control command, when determining whether to execute the coarse-grained control command in this step, it is only necessary to perform the judgment based on the driving environment information that affects the coarse-grained control command execution. That is, in this step, it is determined whether or not to execute the coarse-grained control command based on the traveling environment information corresponding to the coarse-grained control command.
  • the driving environment information corresponding to the coarse-grained control instruction may include driving environment information within a preset range in front of the self-driving vehicle; when the coarse-grained control instruction includes the lane changing direction, The driving environment information corresponding to the granularity control instruction may include driving environment information within a preset range of the self-driving vehicle on the side of the lane changing direction.
  • the second driving environment information may include lane information corresponding to the coarse-grained control instruction in the driving lane of the self-driving vehicle, and the vehicle information corresponding to the coarse-grained control instruction within the preset distance of the self-driving vehicle, the self-driving vehicle At least one of the road surface information corresponding to the coarse-grained control command within the preset distance.
  • the lane information may include a maximum speed limit of the lane and a width of the lane, etc.
  • the information of the vehicle may include the number of vehicles, the direction of travel of the vehicle, and the distance between the vehicle and the self-driving vehicle, etc.
  • the information of the road surface may include the isolation facility on the road surface. Information and information on obstacles on the road, etc.
  • the second driving environment information may be acquired in real time in this step, or may be determined from the first driving environment information acquired in step 101.
  • the fine-grained control command corresponding to the coarse-grained control command is determined according to the lane information and the traveling state information of the self-driving vehicle.
  • the fine-grained control command may be used to control driving parameters of the self-driving vehicle, and may include a steering wheel angle, a throttle size, a braking force, and the like.
  • the lane information may include information of a lane corresponding to the coarse-grained control command in the road on which the self-driving vehicle travels, and the road on which the self-driving vehicle travels includes a lane in which the self-driving vehicle travels and a lane in which the self-driving vehicle allows lane change, and the road on which the self-driving vehicle travels
  • the medium coarse-grained control command corresponding lane may indicate a lane in which the self-driving vehicle is to be driven to execute the coarse-grained control command.
  • the coarse-grained control command corresponding to the lane in the road on which the self-driving vehicle travels may be the lane in which the self-driving vehicle is traveling straight; the coarse-grained control command includes the road driven by the self-driving vehicle when the direction of the lane change is included
  • the medium coarse-grained control command corresponding lane may be a lane that the self-driving vehicle travels when changing lanes.
  • the lane information and the driving state information of the self-driving vehicle in this step may be acquired in real time, or may be determined from the driving state information acquired in step 101 and the first driving environment information.
  • step 104 the fine-grained control command determined in step 104 is output, and the corresponding components of the self-driving vehicle are controlled to achieve the purpose of automatic driving of the vehicle.
  • the coarse-grained control instruction is first calculated by the driving state information, the first driving environment information, and the control instruction decision model, and then it is determined whether the coarse-graining can be performed based on the second driving environment information.
  • the control command that is, before determining the driving parameter of the self-driving vehicle, first calculating the coarse-grained control command to calculate the driving mode of the self-driving vehicle, and when determining that the coarse-grained control command can be executed, converting the coarse-grained control command into a fine
  • the granularity control instruction can avoid outputting unnecessary or erroneous control commands during automatic vehicle control, avoiding the occurrence of non-conforming natural driving behavior or transmitting safety accidents in the self-driving vehicle; meanwhile, in the embodiment of the present invention, it is determined that coarse-grained control can be performed.
  • the coarse-grained control command is converted into a fine-grained control command based on the driving state information of the self-driving vehicle and the lane information corresponding to the coarse-grained control command, that is, the self-driving vehicle and the lane information are fully considered to determine suitable control.
  • Fine-grained control commands for self-driving vehicles improve the accuracy of fine-grained control commands and the accuracy of autonomous vehicle control, avoiding the occurrence of non-compliance with natural driving behavior or sending safety accidents in autonomous vehicles, and reducing the safety hazards of self-driving vehicles , Further improve the automatic driving vehicle safety and comfort.
  • FIG. 2 is a schematic flow diagram of a method of determining fine-grained control instructions in accordance with an embodiment of the present invention.
  • the coarse-grained control command calculated in step 102 may include straight-line, and the lane information in step 104 includes a preset desired speed of the lane in which the self-driving vehicle is traveling straight, and the vehicle is automatically driven.
  • the driving state information includes the current speed of the self-driving vehicle; then when it is determined that the coarse-grained control instruction is executed, as shown in FIG. 2, the step 104 may be specifically performed as the step 1041.
  • the self-driving vehicle can preset a corresponding desired speed for different driving lanes while driving, and the preset desired speed cannot be greater than the maximum speed limit of the lane in order to comply with the driving rules of the vehicle and ensure driving safety.
  • the speed of the self-driving vehicle is controlled based on the current speed of the self-driving vehicle and the preset desired speed of the lane in which it travels straight, that is, the fine-grained control command is calculated to enable safe driving.
  • Controlling the speed of the vehicle is usually achieved by controlling the throttle or brake of the vehicle.
  • the fine-grained control instruction may include the throttle size of the self-driving vehicle.
  • the throttle size When determining the throttle size, it is necessary to first determine the magnitude relationship between the current speed of the self-driving vehicle and the preset desired speed of the lane in which it travels straight.
  • the driving speed of the self-driving vehicle When the current speed is greater than the preset desired speed, the driving speed of the self-driving vehicle is too large, so the throttle can be set to the minimum state at this time, and the throttle size is determined to be zero, so as to reduce the driving speed of the self-driving vehicle as soon as possible;
  • the throttle size is determined based on the difference between the preset desired speed and the current speed and the preset throttle control coefficient to achieve precise control of the throttle. Ensure the safety of self-driving vehicles.
  • the throttle size of the self-driving vehicle may be calculated based on the first preset formula, wherein the first preset formula is as shown in Equation 1.
  • Throttle size preset throttle control coefficient ⁇ (preset desired speed - current speed + preset value) (1)
  • the preset desired speed represents a preset desired speed of the lane in which the self-driving vehicle is traveling straight
  • the preset desired speed of the driving lane when the self-driving vehicle goes straight is the lane of the driving lane when the self-driving vehicle goes straight.
  • the preset desired speed in Equation 1 can be replaced with the value of the highest speed limit.
  • the current speed in Equation 1 represents the current speed of the self-driving vehicle.
  • the throttle size of the self-driving vehicle can be calculated by the above formula 1. Generally, when the throttle size is set to 1, the throttle maximum state is indicated, and when the throttle size is 0, the throttle minimum state is indicated, so the throttle size is taken.
  • the range may be greater than or equal to 0 and less than or equal to 1.
  • the preset value can be set according to the actual application scenario and the specific performance of the vehicle. Generally, the preset value can be equal to 1.
  • the fine-grained control command may include the braking force of the self-driving vehicle.
  • the self-driving vehicle When determining the braking force, it is necessary to first determine the magnitude relationship between the current speed of the self-driving vehicle and the preset desired speed of the lane in which it travels straight. When the current speed is less than the preset desired speed, the driving speed of the self-driving vehicle is normal. At this time, the self-driving vehicle may not brake, so the brake can be set to the minimum state, that is, the braking force is determined to be zero; the current speed is not less than the preset expectation. At the speed, it means that the driving speed of the self-driving vehicle is too large.
  • the speed of the self-driving vehicle needs to be reduced by the brake, so the braking force is determined based on the difference between the preset desired speed and the current speed and the preset braking control coefficient.
  • the braking force of the self-driving vehicle may be calculated based on the second preset formula, wherein the second preset formula is as shown in Formula 2.
  • Brake force preset brake control coefficient ⁇ (current speed - preset desired speed) (2)
  • the preset desired speed represents a preset desired speed of the lane in which the self-driving vehicle is traveling straight
  • the preset desired speed of the driving lane when the self-driving vehicle goes straight is the lane of the driving lane when the self-driving vehicle goes straight.
  • the preset desired speed in Equation 2 can be replaced with the value of the highest speed limit.
  • the current speed in Equation 2 represents the current speed of the self-driving vehicle.
  • the braking force of the self-driving vehicle can be calculated by the above formula 2. Generally, when the braking force is set to 1, the maximum braking force is indicated, and when the braking force is 0, the braking force is the minimum state, so the braking is performed.
  • the value of the velocity may be a value greater than or equal to 0 and less than or equal to 1.
  • the preset throttle control coefficient in Equation 1 and the preset brake control coefficient in Equation 2 can be determined according to the performance of the self-driving vehicle, and the corresponding preset throttle control coefficient and preset for different brands and models of vehicles.
  • the brake control coefficients can be different.
  • the preset brake control coefficient can be 0.1 and the preset throttle control coefficient can be 0.2.
  • the steering wheel angle of the self-driving vehicle does not need to be controlled, and only the traveling speed of the self-driving vehicle can be controlled, so according to the current speed of the self-driving vehicle and the self-driving vehicle
  • the preset desired speed of the straight lane determines the fine-grained control command for the self-driving vehicle, improves the accuracy of the fine-grained control command and the accuracy of the self-driving vehicle control, and avoids the natural driving behavior such as the left-hand swing of the self-driving vehicle.
  • FIG. 3 is a schematic flow chart of still another method for determining a fine-grained control instruction according to an embodiment of the invention.
  • the coarse-grained control command calculated in step 102 may include a lane change direction.
  • step 104 may be specifically executed. It is step 1042 and step 1043.
  • the implementation manner of the simulated lane changing driving path is not limited.
  • the self-driving vehicle when the coarse-grained control command includes the direction change direction, it is stated that the self-driving vehicle needs to perform lane change on the corresponding side of the lane change direction.
  • the lane change travel path of the self-driving vehicle is first simulated, that is, in the automatic manner. Before driving the vehicle to actually change lanes, first determine the lane change route, and then determine the fine-grained control command of the self-driving vehicle during the lane change based on the predetermined lane change route and lane information, thereby improving the accuracy of the fine-grained control command. And the accuracy of the self-driving vehicle control, thereby improving the safety of the self-driving vehicle when changing lanes.
  • the fine-grained control instruction may include a steering wheel angle of the self-driving vehicle.
  • the lane information in the step 1043 includes the lane width of the lane corresponding to the lane change travel route and the lane center line of the lane corresponding to the lane change travel route; step 1043 may be specifically performed as: determining the driving direction of the self-driving vehicle and the automatic driving in the lane change travel route At least one steering angle between the straight-direction direction of the current lane of the vehicle; in the lane-changing driving path, determining the target lane width of the lane of the lane of the self-driving vehicle corresponding to the steering angle according to the lane width of the lane corresponding to the lane-changing driving route; In the driving path, the target distance between the corresponding position of the steering angle and the lane center line of the associated lane is determined according to the lane center line of the lane corresponding to the lane change driving path; based on the third prese
  • the simulated lane-changing driving path is a curved path of a section of the self-driving vehicle, and in the embodiment of the present invention, in order to facilitate the control of the self-driving vehicle, the continuous driving process of the self-driving vehicle in the lane changing driving path may be adopted. Discrete into at least one time point and calculate fine-grained control commands of the self-driving vehicle at these points in time to achieve control of the entire lane-changing travel process of the self-driving vehicle.
  • the at least one position of the self-driving vehicle in the lane change travel path can be determined, and then the driving direction of the self-driving vehicle and the current lane of the self-driving vehicle can be simulated to be straight ahead during the lane change driving.
  • At least one steering angle between the directions determines at least one steering angle between the direction in which the self-driving vehicle travels in the lane change travel path and the direction in which the self-driving vehicle is currently traveling.
  • these steering angles need to be converted to the steering wheel angle of the self-driving vehicle.
  • the steering angle corresponds to the position in the lane change travel path, that is, the position of the self-driving vehicle in the lane change travel path corresponding to the steering angle, thereby determining the self-driving vehicle.
  • the lane center line of the lane to which the steering angle corresponds to the position of the self-driving vehicle in the lane change travel path and the self-driving vehicle may be determined according to the lane center line of the lane corresponding to the lane change travel path.
  • Steering wheel angle [steering angle - target distance / (target lane width ⁇ first corner factor)] ⁇ second corner factor
  • the steering angle represents an angle between the driving direction of the self-driving vehicle and the straight traveling direction of the self-driving vehicle
  • the target distance indicates that the steering angle corresponding position of the self-driving vehicle in the lane changing traveling path and the self-driving vehicle are traveling on the lane change
  • the steering angle in the path corresponds to the distance between the lane centerlines of the lane to which the position belongs
  • the target lane width indicates the target lane width of the lane to which the self-driving vehicle is in the corresponding position of the steering angle in the lane change travel path.
  • the first rotation angle coefficient and the second rotation angle coefficient may be determined according to the performance of the self-driving vehicle, and the corresponding first rotation angle coefficient and the second rotation angle coefficient may be different for different brands and models of vehicles.
  • the first corner factor may take a value of 0.541
  • the first corner factor may take a value of 0.4.
  • the steering wheel angle can be set to -1 when the steering wheel is rotated to the maximum angle of one side, and the steering wheel angle is set to 1 when the steering wheel is rotated to the maximum angle of the other side, the range of the steering wheel angle can be It is greater than or equal to -1 and less than or equal to 1.
  • the steering wheel angle of the self-driving vehicle needs to be controlled. Therefore, based on the pre-simulated lane change travel route, it can be determined that the self-driving vehicle needs to turn when changing lanes. And the steering angle, and then determining the rotation angle of the steering wheel at the respective steering angles in combination with the information of the respective steering angles corresponding to the lanes in the lane-changing driving path, that is, determining the angle at which the steering wheel needs to be rotated when the self-driving vehicle changes lanes, thereby Accurate control of the steering wheel ensures safe driving of the self-driving vehicle.
  • the traveling speed of the self-driving vehicle can also be controlled.
  • the continuous running process of the self-driving vehicle in the lane changing traveling path is discretized into at least one time point, and at least one steering angle is determined.
  • the traveling speed of the self-driving vehicle can be controlled by determining the speed of each of the at least one steering angle in the lane change travel path of the self-driving vehicle.
  • the fine-grained control command also includes the throttle size of the self-driving vehicle.
  • the lane information in step 104 may further include a preset desired speed of the lane corresponding to the lane change travel path.
  • the following process may also be performed: determining, in the lane changing driving path, the current speed of the automatically driving vehicle at the corresponding position of the steering angle based on the driving state information of the automatically driving vehicle In the lane changing driving path, determining, according to the preset desired speed of the lane corresponding to the lane change driving path, the target preset desired speed of the lane of the autopilot vehicle at the corresponding position of the steering angle; when the current speed is greater than the target preset desired speed, the throttle The size is equal to zero; when the current speed is not greater than the target preset desired speed, the throttle size of the auto-driving vehicle at the corresponding position of the steering angle is calculated based on the fourth preset formula.
  • each of the at least one steering angle can determine the corresponding throttle size. Firstly, in the lane changing driving path, the current speed of the autopilot vehicle at the corresponding position of the steering angle and the target preset desired speed of the lane of the autopilot vehicle at the corresponding position of the steering angle are determined, and the throttle size can be determined according to the calculation of Equation 4.
  • Throttle size preset throttle control coefficient ⁇ (target preset desired speed - current speed + preset value) (4)
  • the target preset desired speed in Formula 4 represents the target preset desired speed of the lane to which the steering angle corresponding position of the self-driving vehicle is in the lane changing traveling path
  • the target preset desired speed is the steering of the self-driving vehicle in the lane changing driving path.
  • the target preset desired speed in Equation 4 can be replaced with the value of the highest speed limit.
  • the current speed in Equation 4 represents the current speed of the corresponding position of the steering angle of the self-driving vehicle in the lane change travel path.
  • the traveling speed can also be controlled.
  • the throttle size is determined to be zero, so as to reduce the driving speed of the self-driving vehicle as soon as possible; when the current speed is not greater than the target preset desired speed, the target preset desired speed and the current speed are determined based on the target.
  • the difference between the difference and the preset throttle control coefficient determines the throttle size to achieve precise control of the throttle while ensuring the safety of the self-driving vehicle.
  • the fine-grained control command may also include the braking force of the self-driving vehicle.
  • the lane information in step 104 may further include a preset desired speed of the lane corresponding to the lane change travel path.
  • a process may also be performed in which, in the lane change travel path, the current state of the autopilot vehicle at the corresponding position of the steering angle is determined based on the travel state information of the self-driving vehicle.
  • each of the at least one steering angle can determine the corresponding braking force. Firstly, in the lane changing driving path, the current speed of the autopilot vehicle at the corresponding position of the steering angle and the target preset desired speed of the lane of the autopilot vehicle at the corresponding position of the steering angle are determined, and the braking force can be determined according to the calculation of Equation 5.
  • Brake force preset brake control coefficient ⁇ (current speed - target preset desired speed) (5)
  • the target preset desired speed in Formula 5 represents the target preset desired speed of the lane to which the steering angle corresponding position of the self-driving vehicle is in the lane changing traveling path
  • the target preset desired speed is the steering of the self-driving vehicle in the lane changing driving path.
  • the target preset desired speed in Equation 5 can be replaced with the value of the highest speed limit.
  • the current speed in Equation 5 represents the current speed of the corresponding position of the steering angle of the self-driving vehicle in the lane change travel path.
  • the traveling speed can also be controlled.
  • the brake may not be performed, so the braking force is adjusted to zero;
  • the target preset is based on the target
  • the difference between the desired speed and the current speed and the preset brake control coefficient determine the braking force to achieve precise control of the brakes and ensure the safety of the self-driving vehicle.
  • the throttle size, the preset throttle control coefficient and the preset value in Equation 4 are the same as the throttle size, the preset throttle control coefficient and the preset value in Equation 1.
  • the value of the braking force and the preset brake control coefficient in Equation 5 is the same as the value of the throttle size and the preset brake control coefficient in Equation 2.
  • the second driving environment information may include a first vehicle distance between the target vehicle and the self-driving vehicle, and the target vehicle indicates the preset distance and coarse-grained control of the self-driving vehicle.
  • the step 103 may be specifically performed to: when the first vehicle distance is greater than the safety distance, determine to execute the coarse-grained control instruction; when the first vehicle distance is not greater than the safety distance, determine that the coarse-grained control instruction is not executed.
  • the target vehicle when the coarse-grained control instruction includes straight-line, the target vehicle represents a preset distance in front of the self-driving vehicle The vehicle within the departure; when the coarse-grained control command includes the lane change direction, the target vehicle represents the vehicle within the preset distance of the self-driving vehicle on the side of the lane change direction. If the vehicle corresponding to the coarse-grained control command within the preset distance of the self-driving vehicle is zero, that is, the vehicle corresponding to the coarse-grained control instruction does not exist within the preset distance of the self-driving vehicle, the first vehicle distance may be determined to be infinite.
  • the self-driving vehicle may discard the coarse-grained control command to maintain the current driving state.
  • the processing described in step 1041 in the embodiment of the present invention may be performed to facilitate real-time control of the traveling speed of the self-driving vehicle.
  • the safety distance may be: the self-driving vehicle runs at the current speed of the self-driving vehicle, and the target vehicle runs at the current speed of the target vehicle, and does not make the between the self-driving vehicle and the target vehicle within the preset time period.
  • the distance of contact That is, the safety distance needs to ensure that the self-driving vehicle travels at the current speed of the self-driving vehicle for a preset period of time, while the target vehicle travels at the current speed of the target vehicle for the same preset time period, during which the self-driving vehicle and the target vehicle do not There will be a collision.
  • whether or not the vehicle corresponding to the coarse-grained control command affects the safety of the self-driving vehicle may be determined by determining whether the vehicle corresponding to the coarse-grained control command within the preset distance of the self-driving vehicle determines The coarse-grained control command determines whether the first vehicle distance between the target vehicle and the self-driving vehicle reaches a safe distance.
  • the coarse-grained control instruction is not executed when the first vehicle distance does not reach the safety distance, and the coarse-grained control instruction is executed when the first vehicle distance reaches the safety distance, which can be effective Improve the safety of self-driving vehicles in automatic driving, avoiding the occurrence of safety accidents caused by the execution of incorrect or unnecessary control commands by the self-driving vehicle.
  • step 106 when the coarse-grained control instruction includes the lane change direction, step 106 may be further performed before step 103: simulating the auto-driving vehicle based on the driving state information of the self-driving vehicle The lane change route in the direction of the road.
  • the second driving environment information at this time includes a second vehicle distance when the target vehicle and the self-driving vehicle are traveling on the lane changing traveling path, and the target vehicle indicates the vehicle corresponding to the coarse grain control command within the preset distance of the autonomous driving vehicle.
  • Step 103 may be specifically performed to determine that the coarse-grained control instruction is executed when the second vehicle distance is greater than the safety distance; and when the second vehicle distance is not greater than the safety distance, determine that the coarse-grained control instruction is not executed.
  • the target vehicle represents the vehicle within the preset distance of the self-driving vehicle on the side of the lane change direction. If the vehicle corresponding to the coarse-grained control command within the preset distance of the self-driving vehicle is zero, that is, the vehicle corresponding to the coarse-grained control instruction does not exist within the preset distance of the self-driving vehicle, the second vehicle distance may be determined to be infinite.
  • the self-driving vehicle may discard the coarse-grained control command to maintain the current driving state.
  • the safety distance may be: the self-driving vehicle runs at the current speed of the self-driving vehicle, and the target vehicle travels at the current speed of the target vehicle, and enables the self-driving vehicle and the preset time period.
  • the lane-changing driving path of the self-driving vehicle may be simulated first, and then the second vehicle distance pair when the target vehicle and the self-driving vehicle are traveling on the lane-changing driving route are Whether to perform coarse-grained control commands for judgment, thereby improving the accuracy of the judgment and the safety of the self-driving vehicle in automatic driving.
  • the embodiment of the present invention before performing step 102, further includes a process of training the control instruction decision model. Specifically, the following steps may be included.
  • the initialization method can be randomly generated.
  • the training parameters include training driving state information and training driving environment information, that is, training driving state information when the self-driving vehicle is in the training state and training driving environment information when the self-driving vehicle is in the training state.
  • the training driving state information may be the same as the content included in the driving state information in step 101, except that the training driving state information is acquired when the self-driving vehicle is in the training state, and the driving state information is obtained when the self-driving vehicle is in the automatic driving state.
  • the training driving environment information may be the same as the content included in the driving environment information in step 101, except that the training driving environment information is acquired when the self-driving vehicle is in the training state, and the driving environment information is obtained when the self-driving vehicle is in the automatic driving state. of.
  • the model parameters need to be updated continuously during the training process until the accuracy or accuracy requirement of the control instruction decision model is reached, and the update mode of the model parameters is determined by calculating whether the coarse-grained training control command satisfies the requirements.
  • the loss function Loss1 can be as shown in Equation 6.
  • v represents the current speed of the self-driving vehicle
  • ⁇ represents the angle between the current driving direction of the self-driving vehicle and the lane where the self-driving vehicle is located
  • Q represents the coarse-grained training control command and the training parameter corresponding to the preset coarse-grained training control command. The degree of matching between.
  • v represents the traveling speed of the self-driving vehicle in the training state.
  • the value of the calculated loss function is compared with a preset condition to determine whether the model parameter in the control instruction decision model has satisfied the requirement, that is, whether the control instruction decision model is trained to be completed.
  • the value of the loss function does not reach the preset condition, it indicates that the model parameters have not met the requirements, so it is necessary to update the model parameters and continue to train the control instruction decision model.
  • the preset condition may be determined according to the requirements of the model parameters or the standards required to control the instruction decision model.
  • the preset condition may include that the value of the loss function is continuously within the preset range reaches a preset threshold.
  • step C can be re-executed according to the updated model parameters, and the updated coarse-grained training control instruction is obtained, and then step D is used to calculate a new loss function value, and then the new loss function value and the preset condition are obtained.
  • steps E and F are executed, that is, the loop is executed until the value of the loss function calculated in step D reaches the preset condition.
  • the corresponding model parameter when the value of the loss function reaches the preset condition is determined as the final model parameter of the control instruction decision model.
  • the value of the calculated loss function has reached a preset condition, indicating that the model parameter can satisfy the requirement, and the control instruction decision model is trained.
  • the corresponding model parameter is Control the model parameters obtained after the instruction decision model is completed.
  • control instruction decision model may be a neural network model
  • the algorithm for training the control instruction decision model may be an actor-critic reinforcement learning algorithm.
  • the driving strategy decision model is the actor network, and the value of Q is calculated by the critics network.
  • the critics network can also be a neural network model, inputting the training parameters obtained in step B and the coarse-grained training control commands calculated in step C, and the output is a Q value, and the parameters of the critic network are initialized in step A, at each step.
  • E updates the model parameters and updates the parameters in the critic network.
  • the update method is not limited, and can be implemented by a back propagation algorithm.
  • the step E before the step E updates the model parameters, it is also required to calculate an update gradient of the model parameters.
  • the updated model parameters may be calculated based on the update gradient, the preset update coefficient, and the model parameters before the update.
  • the updated model parameters can be calculated according to Equation 7.
  • ⁇ represents the updated model parameter
  • ⁇ 1 represents the model parameter before the update
  • ⁇ represents the preset update coefficient. Indicates the update gradient.
  • the update gradient of the calculation model parameter may be specifically executed.
  • the process is as follows: when it is determined that the coarse-grained training control instruction is executed, the update gradient is calculated by the first preset relationship; when it is determined that the training driving strategy is not executed, the update gradient is calculated by the second preset relationship.
  • the first preset relationship may include: the update gradient is equal to the partial derivative value of the first loss function to the model parameter, and the first loss function is as shown in Equation 6.
  • the second preset relationship may include: the update gradient is equal to the partial derivative value of the second loss function to the model parameter, and the second loss function is as shown in Equation 8.
  • Equation 8 The meanings of the parameters in Equation 8 are the same as those in Equation 6, and are not described here.
  • Equation 7 When it is determined that the coarse-grained training control instruction is executed, in Equation 7 When it is determined that the coarse-grained training control instruction is not executed, Equation 7
  • FIG. 4 is a schematic block diagram of a control device 200 for a vehicle in accordance with an embodiment of the present invention. As shown in FIG. 4, the apparatus 200 includes:
  • the acquiring unit 201 is configured to acquire driving state information of the self-driving vehicle and first driving environment information of the self-driving vehicle when the autonomous driving vehicle is in an automatic driving state;
  • the calculating unit 202 is configured to calculate a coarse-grained control instruction of the self-driving vehicle based on the driving state information, the first driving environment information, and the control instruction decision model, and the coarse-grained control instruction is used to control the driving mode of the self-driving vehicle, and the control instruction decision model It is trained based on the training driving state information of the self-driving vehicle in the training state and the training driving environment information when the self-driving vehicle is in the training state;
  • the determining unit 203 is configured to determine, according to the second driving environment information, whether to execute the coarse-grained control instruction, where the second driving environment information includes driving environment information corresponding to the coarse-grained control instruction when the self-driving vehicle is in the automatic driving state;
  • the determining unit 204 is configured to, when determining to execute the coarse-grained control instruction, determine a fine-grained control instruction corresponding to the coarse-grained control instruction according to the lane information and the driving state information of the self-driving vehicle, and the fine-grained control instruction is used to control the self-driving vehicle
  • the driving parameter, the lane information includes information corresponding to the lane of the coarse-grained control command in the road on which the self-driving vehicle is traveling;
  • the output unit 205 is configured to output a fine-grained control instruction.
  • the control device 200 first calculates the coarse-grained control command by using the driving state information, the first driving environment information, and the control command decision model, and then determines whether the coarse-grained control command can be executed based on the second driving environment information, that is, Before determining the driving parameters of the self-driving vehicle, the coarse-grained control command is calculated to calculate the driving mode of the self-driving vehicle, and when it is determined that the coarse-grained control command can be executed, the coarse-grained control command is converted into the fine-grained control command.
  • the control device determines that the coarse-grained control command can be executed.
  • the coarse-grained control command is converted into a fine-grained control command based on the driving state information of the self-driving vehicle and the lane information corresponding to the coarse-grained control command, that is, fully considering the self-driving state of the self-driving vehicle and the lane information to determine that it is suitable for controlling the automatic driving.
  • Vehicle Fine-grained control commands to improve the accuracy of fine-grained control commands and the accuracy of autonomous vehicle control, to avoid auto-driving vehicles that do not conform to natural driving behavior or to send safety incidents, to reduce safety hazards of autonomous vehicles, and to further improve automation Driving the safety and comfort of the vehicle.
  • the coarse-grained control instruction includes straight line
  • the lane information includes a preset desired speed of the lane in which the self-driving vehicle travels straight, and the driving state information of the self-driving vehicle includes the current speed of the self-driving vehicle;
  • the determining unit 204 is specifically configured to determine the fine-grained control instruction based on the preset desired speed and the current speed.
  • the fine-grained control command includes the throttle size of the self-driving vehicle
  • the determining unit 204 is specifically configured to:
  • the throttle size is equal to zero
  • the throttle size of the self-driving vehicle is calculated based on the first preset formula, wherein the first preset formula includes:
  • Throttle size preset throttle control coefficient ⁇ (preset desired speed - current speed + preset value).
  • the fine-grained control command includes the braking force of the self-driving vehicle
  • the determining unit 204 is specifically configured to:
  • the braking force is equal to zero
  • the braking force of the self-driving vehicle is calculated based on the second preset formula, wherein the second preset formula includes:
  • Brake force preset brake control coefficient ⁇ (current speed - preset desired speed).
  • the coarse-grained control instruction includes a lane change direction
  • the determining unit 204 is specifically configured to:
  • a fine-grained control command of the self-driving vehicle is determined based on the lane change travel route and the lane information.
  • the fine-grained control instruction includes a steering wheel angle
  • the lane information includes a lane width corresponding to the lane of the lane change travel route and a lane centerline corresponding to the lane of the lane change travel route;
  • the determining unit 204 is specifically configured to:
  • the steering wheel angle is calculated based on the third preset formula, the steering angle, the target lane width, and the target distance, wherein the third preset formula includes:
  • Steering wheel angle [steering angle - target distance / (target lane width ⁇ first corner factor)] ⁇ second corner factor.
  • the fine-grained control command also includes the throttle size of the self-driving vehicle
  • the lane information further includes a preset desired speed of the lane corresponding to the lane change driving path;
  • the determining unit 204 is specifically configured to:
  • the throttle size is equal to zero
  • the self-driving vehicle is calculated based on the fourth preset formula
  • the throttle size at the corresponding position of the steering angle wherein the fourth preset formula includes:
  • Throttle size preset throttle control coefficient ⁇ (target preset desired speed - current speed + preset value).
  • the fine-grained control command also includes the braking force of the self-driving vehicle
  • the lane information further includes a preset desired speed of the lane corresponding to the lane change driving path;
  • the determining unit 204 is specifically configured to:
  • the braking force is equal to zero
  • the braking force of the self-driving vehicle at the corresponding position of the steering angle is calculated based on the fifth preset formula, wherein the fifth preset formula includes:
  • Brake force preset brake control coefficient ⁇ (current speed - target preset desired speed).
  • the first driving environment information includes lane information of a lane in which the self-driving vehicle travels, information of the vehicle within the preset distance of the self-driving vehicle, and at least one of the information of the road surface within the preset distance of the self-driving vehicle;
  • the second driving environment information includes lane information corresponding to the coarse-grained control instruction in the driving lane of the self-driving vehicle, and the vehicle information corresponding to the coarse-grained control instruction within the preset distance of the self-driving vehicle, and the preset distance and coarse granularity of the self-driving vehicle At least one of the road surface information corresponding to the control command.
  • the second driving environment information includes a first vehicle distance between the target vehicle and the self-driving vehicle, and the target vehicle represents a vehicle corresponding to the coarse-grained control instruction within the preset distance of the self-driving vehicle;
  • the determining unit 203 is specifically configured to:
  • FIG. 5 is a schematic block diagram of a control device 200 of a vehicle according to still another embodiment of the present invention.
  • the coarse-grained control instruction includes a lane change direction
  • control device 200 may further include:
  • the simulation unit 206 is configured to simulate, according to the driving state information of the self-driving vehicle, a lane changing driving path of the self-driving vehicle in the lane changing direction;
  • the second driving environment information includes a second vehicle distance when the target vehicle and the self-driving vehicle are traveling on the lane changing driving path, and the target vehicle indicates the vehicle corresponding to the coarse grain control instruction within the preset distance of the autonomous driving vehicle;
  • the determining unit 203 is specifically configured to:
  • control device 200 may further include:
  • An initialization unit 207 configured to initialize model parameters of the control instruction decision model
  • the obtaining unit 201 is further configured to acquire training parameters, where the training parameters include training driving state information and training driving environment information;
  • the calculating unit 202 is further configured to calculate coarse-grained training of the self-driving vehicle according to the model parameters and the training parameters. a control instruction; and a value for calculating a loss function according to the coarse-grained training control instruction;
  • the updating unit 208 is configured to update the model parameter when the value of the loss function does not reach the preset condition
  • the calculating unit 202 is further configured to calculate an updated coarse-grained training control instruction of the self-driving vehicle according to the updated model parameter and the training parameter, and recalculate the value of the loss function until the value of the loss function reaches a preset condition;
  • the determining unit 204 is further configured to determine the corresponding model parameter as the final model parameter of the control instruction decision model when the value of the loss function reaches a preset condition.
  • loss function Loss1 includes:
  • Loss1
  • v represents the current speed of the self-driving vehicle
  • ⁇ represents the angle between the current driving direction of the self-driving vehicle and the lane where the self-driving vehicle is located
  • Q represents the coarse-grained training control command and the training parameter corresponding to the preset coarse-grained training control command. The degree of matching between.
  • calculation unit 202 is further configured to calculate an update gradient of the model parameters
  • the updating unit 208 is specifically configured to calculate the updated model parameters based on the update gradient, the preset update coefficient, and the model parameters before the update.
  • the determining unit 203 is further configured to determine whether to execute the coarse-grained training control instruction
  • the calculating unit 202 is specifically configured to:
  • the control device 200 of the vehicle may correspond to an execution subject in a control method of a vehicle according to an embodiment of the present invention, and the above-described and other operations and/or functions of respective modules in the control device 200 of the vehicle are respectively The respective processes in the control method of the vehicle are implemented, and are not described herein again for the sake of brevity.
  • FIG. 6 is a schematic block diagram of a control device 300 of a vehicle according to an embodiment of the present invention.
  • the device 300 includes a processor 301 and a memory 302 summary 303.
  • the bus 303 is used to connect the processor 301 and the memory 302, so that the processor 301 and the memory 302 communicate with each other through a bus 303.
  • the executed program code, the processor 301 runs the program corresponding to the executable program code by reading the executable program code stored in the memory 302.
  • the processor 301 is further configured to execute a control method of the vehicle; wherein the control method of the vehicle includes:
  • the coarse-grained control command of the vehicle is used to control the driving mode of the self-driving vehicle.
  • the control command decision model is based on the training driving state information of the self-driving vehicle in the training state and the training driving when the self-driving vehicle is in the training state. Obtained by environmental information training;
  • the second driving environment information includes driving environment information corresponding to the coarse-grained control instruction when the self-driving vehicle is in the automatic driving state
  • the fine-grained control instruction corresponding to the coarse-grained control instruction is determined according to the lane information and the driving state information of the self-driving vehicle, and the fine-grained control instruction is used to control the driving parameter of the self-driving vehicle, the lane information
  • the control device 300 of the vehicle according to an embodiment of the present invention may correspond to an execution subject in a control method of a vehicle according to an embodiment of the present invention, and the above-described and other operations and/or functions of respective modules in the control device 300 of the vehicle are respectively In order to implement the respective corresponding processes in the control method of the vehicle, for brevity, no further details are provided herein.
  • the above embodiments it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof.
  • software it may be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the computer program instructions When the computer program instructions are loaded and executed on a computer, the processes or functions described in accordance with embodiments of the present invention are generated in whole or in part.
  • the computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable device.
  • the computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be from a website site, computer, server or data center Transfer to another website site, computer, server, or data center by wire (eg, coaxial cable, fiber optic, digital subscriber line (DSL), or wireless (eg, infrared, wireless, microwave, etc.).
  • the computer readable storage medium can be any available media that can be accessed by a computer or a data storage device such as a server, data center, or the like that includes one or more available media.
  • the usable medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (such as an olid state disk (SSD)) or the like.
  • a magnetic medium eg, a floppy disk, a hard disk, a magnetic tape
  • an optical medium eg, a DVD
  • a semiconductor medium such as an olid state disk (SSD)

Landscapes

  • Engineering & Computer Science (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Automation & Control Theory (AREA)
  • Human Computer Interaction (AREA)
  • Evolutionary Computation (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Artificial Intelligence (AREA)
  • Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Medical Informatics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Health & Medical Sciences (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
  • Traffic Control Systems (AREA)
  • Steering Control In Accordance With Driving Conditions (AREA)

Abstract

一种车辆的控制方法、装置及设备。该方法包括：在自动驾驶车辆处于自动驾驶状态时，获取自动驾驶车辆的行驶状态信息和自动驾驶车辆的第一行驶环境信息（101）；基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车辆的粗粒度控制指令（102）；根据第二行驶环境信息判断是否执行粗粒度控制指令（103）；当判定执行粗粒度控制指令时，根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令（104）。该车辆的控制方法、装置及设备能够解决自动驾驶控制决策系统输出的控制指令给自动驾驶的车辆带来安全隐患的问题。

Description

车辆的控制方法、装置及设备 技术领域
本发明涉及通信技术领域，尤其涉及一种车辆的控制方法、装置及设备。
背景技术
自动驾驶系统是集自动控制、体系结构、人工智能、视觉计算等众多技术于一体的智能系统。自动驾驶系统中控制决策系统根据感知系统解析出的信息可以确定出自动驾驶的策略，然后输出相应的控制指令，以实现规划车辆行车路线并控制车辆自动行驶的目的。
控制决策系统在确定自动驾驶策略时需要考虑到汽车行驶的安全性和舒适性，需要在尽快到达目的地的同时保证驾驶过程的安全。因此，自动驾驶系统中对控制决策系统控制车辆自动驾驶时的安全性和可靠性有着严格要求。
现有技术中，为了保证车辆自动驾驶时的安全性和可靠性，控制决策系统通常基于自动驾驶车辆在行驶时的各项参数，利用学习算法计算并输出自动驾驶车辆的控制指令，如方向盘角度、油门大小、刹车力度等等，通过输出的控制指令对自动驾驶车辆进行控制。但是，在控制决策系统通过控制指令控制车辆自动驾驶时，自动驾驶车辆经常会出现不符合自然驾驶行为的情况，例如，自动驾驶车辆出现频繁的左右摆动，甚至导致安全事故发生，因此会给自动驾驶的车辆带来安全隐患。
发明内容
本发明实施例提供了一种车辆的控制方法、装置及设备，能够解决现有技术中通过控制指令控制车辆自动驾驶时给自动驾驶的车辆带来安全隐患的问题。
第一方面，本发明实施例提供了一种车辆的控制方法，包括：
在自动驾驶车辆处于自动驾驶状态时，获取自动驾驶车辆的行驶状态信息和自动驾驶车辆的第一行驶环境信息；
基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车辆的粗粒度控制指令，粗粒度控制指令用于控制自动驾驶车辆的行驶方式，控制指令决策模型为基于自动驾驶车辆在训练状态时的训练行驶状态信息和自动驾驶车辆在训练状态时的训练行驶环境信息训练得到的；
根据第二行驶环境信息判断是否执行粗粒度控制指令，第二行驶环境信息包括在自动驾驶车辆处于自动驾驶状态时，与粗粒度控制指令对应的行驶环境信息；
当判定执行粗粒度控制指令时，根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令，细粒度控制指令用于控制自动驾驶车辆的行驶参数，车道信息包括自动驾驶车辆所行驶道路中粗粒度控制指令对应车道的信息；
输出细粒度控制指令。
本发明实施例中，通过行驶状态信息、第一行驶环境信息和控制指令决策模型先计算出粗粒度控制指令，然后基于第二行驶环境信息来判定是否可以执行粗粒度控制指令，即在确定控制自动驾驶车辆的行驶参数之前，先计算出粗粒度控制指令计算出自动驾驶车辆的行驶方式，当判定可以执行粗粒度控制指令时，再将粗粒度控制指令转换为细粒度控制指令，从而能够避免在车辆自动控制时输出不必要或错误的控制指令，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故；同时，本发明实施例中在判定可以执行粗粒度控制指令后，基于自动驾驶车辆的行驶状态信息和粗粒度控制指令对应的车道信息将粗粒度控制指令转换为细粒度控制指令，即充分考虑自动驾驶车辆行驶时的自身状况和车道信息确定出适合控制自动驾驶车辆的细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故，从降低自动驾驶的车辆的安全隐患，进一步提高自动驾驶车辆的安全性和舒适性。
结合第一方面，在第一方面的第一种可能的实施方式中，粗粒度控制指令包括直行；
车道信息包括自动驾驶车辆直行时所行驶车道的预设期望速度，自动驾驶车辆的行驶状态信息包括自动驾驶车辆的当前速度；
根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令，包括：
基于预设期望速度和当前速度，确定细粒度控制指令。
本实施方式中，粗粒度控制指令包括直行时，不需要对自动驾驶车辆的方向盘转角进行控制，可以只控制自动驾驶车辆的行驶速度，所以根据自动驾驶车辆的当前的速度和自动驾驶车辆在直行车道的预设期望速度来对自动驾驶车辆来确定细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，避免自动驾驶车辆出现左右摆动等不符合自然驾驶行为。
结合第一方面或上述可能的实施方式，在第一方面的第二种可能的实施方式中，细粒度控制指令包括自动驾驶车辆的油门大小；
基于预设期望速度和当前速度，确定细粒度控制指令，包括：
当当前速度大于预设期望速度时，油门大小等于零；
当当前速度不大于预设期望速度时，基于第一预设公式计算自动驾驶车辆的油门大小，其中，第一预设公式包括：
油门大小＝预设油门控制系数×(预设期望速度-当前速度+预设值)。
本实施方式中，当前速度大于预设期望速度时，说明自动驾驶车辆的行驶速度过大，所以此时将油门大小确定为零，以便于尽快降低自动驾驶车辆的行驶速度；当前速度不大于预设期望速度时，说明自动驾驶车辆的行驶速度正常，此时基于预设期望速度和当前速度之间的差值和预设油门控制系数确定油门大小，以实现对油门的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中预设期望速度不大于自动驾驶车辆直行时所行驶车道的最高限速。
结合第一方面或上述可能的实施方式，在第一方面的第三种可能的实施方式中， 细粒度控制指令包括自动驾驶车辆的刹车力度；
基于预设期望速度和当前速度，确定细粒度控制指令，包括：
当当前速度小于预设期望速度时，刹车力度等于零；
当当前速度不小于预设期望速度时，基于第二预设公式计算自动驾驶车辆的刹车力度，其中，第二预设公式包括：
刹车力度＝预设刹车控制系数×(当前速度-预设期望速度)。
本实施方式中，当前速度小于预设期望速度时，说明自动驾驶车辆的行驶速度正常，此时可以不进行刹车，所以将刹车力度调整为零；当前速度不小于预设期望速度时，说明自动驾驶车辆的行驶速度过大，此时需要通过刹车来控制自动驾驶车辆的速度，所以基于预设期望速度和当前速度之间的差值和预设刹车控制系数确定刹车力度，以实现对刹车的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中预设期望速度不大于自动驾驶车辆直行时所行驶车道的最高限速。
结合第一方面或上述可能的实施方式，在第一方面的第四种可能的实施方式中，粗粒度控制指令包括换道方向；
根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令，包括：
基于自动驾驶车辆的行驶状态信息，模拟自动驾驶车辆在换道方向的换道行驶路径；
根据换道行驶路径和车道信息，确定自动驾驶车辆的细粒度控制指令。
本实施方式中，粗粒度控制指令包括换道方向时，控制自动驾驶车辆向换道方向侧进行换道，此时首先模拟出自动驾驶车辆的换道行驶路径，即在自动驾驶车辆实际进行换道之前，先确定出换道路径，然后再基于预先确定的换道路径和车道信息确定自动驾驶车辆在换道时的细粒度控制指令，从而提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，进而提高自动驾驶车辆换道时的安全性。
结合第一方面或上述可能的实施方式，在第一方面的第五种可能的实施方式中，细粒度控制指令包括方向盘转角；
车道信息包括换道行驶路径对应车道的车道宽度和换道行驶路径对应车道的车道中线；
根据换道行驶路径和车道信息，确定自动驾驶车辆的细粒度控制指令，包括：
确定在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度；
在换道行驶路径中，根据换道行驶路径对应车道的车道宽度确定自动驾驶车辆在转向角度对应位置的所属车道的目标车道宽度；
在换道行驶路径中，根据换道行驶路径对应车道的车道中线确定自动驾驶车辆在转向角度对应位置与所属车道的车道中线之间的目标距离；
基于第三预设公式、转向角度、目标车道宽度和目标距离计算方向盘转角，其中，第三预设公式包括：
方向盘转角＝[转向角度-目标距离/(目标车道宽度×第一转角系数)]×第二转角 系数。
本实施方式中，由于自动驾驶车辆的换道行驶路径为曲线，需要对自动驾驶车辆的方向盘转角进行控制，所以基于预先模拟的换道行驶路径可以确定出自动驾驶车辆在换道时离散的需要转向的位置和转向角度，然后在结合换道行驶路径中各转向角度对应车道的信息确定出方向盘在对应各转向角度时的转角，即确定出自动驾驶车辆在换道时方向盘需要转动的角度，从而实现对方向盘的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中预设期望速度不大于自动驾驶车辆直行时所行驶车道的最高限速。方向盘转角的取值范围可以为大于等于-1且小于等于1。
结合第一方面或上述可能的实施方式，在第一方面的第六种可能的实施方式中，细粒度控制指令还包括自动驾驶车辆的油门大小；
车道信息包括换道行驶路径对应车道的预设期望速度；
确定在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度之后，还包括：
在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置的当前速度；
在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度；
当当前速度大于目标预设期望速度时，油门大小等于零；
当当前速度不大于目标预设期望速度时，基于第四预设公式计算自动驾驶车辆在转向角度对应位置的油门大小，其中，第四预设公式包括：
油门大小＝预设油门控制系数×(目标预设期望速度-当前速度+预设值)。
本实施方式中，在自动驾驶车辆换道时，还可以对行驶速度进行控制。在当前速度大于目标预设期望速度时，将油门大小确定为零，以便于尽快降低自动驾驶车辆的行驶速度；当前速度不大于目标预设期望速度时，基于目标预设期望速度和当前速度之间的差值和预设油门控制系数确定油门大小，以实现对油门的精确控制，同时保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中目标预设期望速度不大于换道行驶路径对应车道的最高限速。
结合第一方面或上述可能的实施方式，在第一方面的第七种可能的实施方式中，细粒度控制指令还包括自动驾驶车辆的刹车力度；
车道信息还包括换道行驶路径对应车道的预设期望速度；
确定在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度之后，还包括：
在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置所属车道的当前速度；
在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置的目标预设期望速度；
当当前速度小于目标预设期望速度时，刹车力度等于零；
当当前速度不小于目标预设期望速度时，基于第五预设公式计算自动驾驶车辆在转向角度对应位置的刹车力度，其中，第五预设公式包括：
刹车力度＝预设刹车控制系数×(当前速度-目标预设期望速度)。
本实施方式中，在自动驾驶车辆换道时，还可以对行驶速度进行控制。当前速度小于目标预设期望速度时，可以不进行刹车，所以将刹车力度调整为零；当前速度不小于目标预设期望速度时，需要通过刹车来控制自动驾驶车辆的速度，所以基于目标预设期望速度和当前速度之间的差值和预设刹车控制系数确定刹车力度，以实现对刹车的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，在上述实施方式中目标预设期望速度不大于换道行驶路径对应车道的最高限速。油门大小的取值范围可以为大于等于0且小于等于1。刹车力度的取值范围可以为大于等于0且小于等于1。预设值的取值可以为1。
结合第一方面或上述可能的实施方式，在第一方面的第八种可能的实施方式中，第一行驶环境信息包括自动驾驶车辆所行驶车道的车道信息，自动驾驶车辆预设距离内车辆的信息，自动驾驶车辆预设距离内路面的信息中至少一项；
第二行驶环境信息包括自动驾驶车辆所行驶车道中与粗粒度控制指令对应的车道信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的路面信息中至少一项。
结合第一方面或上述可能的实施方式，在第一方面的第九种可能的实施方式中，第二行驶环境信息包括目标车辆与自动驾驶车辆之间的第一车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆；
根据第二行驶环境信息判断是否执行粗粒度控制指令，包括：
当第一车辆距离大于安全距离时，判定执行粗粒度控制指令；
当第一车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
本实施方式中，在计算出粗粒度控制指令后，从安全角度出发，可以通过判断自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆是否会影响自动驾驶车辆的安全来确定是否执行粗粒度控制指令，即判断目标车辆与自动驾驶车辆之间的第一车辆距离是否达到安全距离。通过对第一车辆距离和安全距离之间大小的判断，可以有效的提高自动驾驶车辆在自动驾驶中的安全性，避免自动驾驶车辆执行错误或不必要的控制指令导致安全事故的发生。
结合第一方面或上述可能的实施方式，在第一方面的第十种可能的实施方式中，粗粒度控制指令包括换道方向；
在根据第二行驶环境信息判断是否执行粗粒度控制指令之前，还包括：
基于自动驾驶车辆的行驶状态信息模拟自动驾驶车辆在换道方向的换道行驶路径；
第二行驶环境信息包括目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆；
根据第二行驶环境信息判断是否执行粗粒度控制指令，包括：
当第二车辆距离大于安全距离时，判定执行粗粒度控制指令；
当第二车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
本实现方式中，粗粒度控制指令包括换道方向时，可以先模拟出自动驾驶车辆的换道行驶路径，然后基于目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离对是否执行粗粒度控制指令进行判断，从而提高判断的准确性和自动驾驶车辆在自动驾驶中的安全性。
具体的，在上述实施例中，安全距离可以为：自动驾驶车辆以自动驾驶车辆的当前速度行驶、且目标车辆以目标车辆的当前速度行驶，在预设时间段内使自动驾驶车辆和目标车辆之间不接触的距离。
结合第一方面或上述可能的实施方式，在第一方面的第十一种可能的实施方式中，在基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车辆的粗粒度控制指令之前，还包括：
初始化控制指令决策模型的模型参数；
获取训练参数，训练参数包括训练行驶状态信息和训练行驶环境信息；
根据模型参数和训练参数计算自动驾驶车辆的粗粒度训练控制指令；
依据粗粒度训练控制指令计算损失函数的值；
当损失函数的值未达到预设条件时，更新模型参数；
根据更新后的模型参数和训练参数计算自动驾驶车辆的更新粗粒度训练控制指令，并重新计算损失函数的值，直到损失函数的值达到预设条件；
将损失函数的值达到预设条件时对应的模型参数确定为控制指令决策模型的最终模型参数。
结合第一方面或上述可能的实施方式，在第一方面的第十二种可能的实施方式中，损失函数Loss1包括：
Loss1＝|vcosα-vsinα-Q|2，
其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预设粗粒度训练控制指令之间匹配的程度。
结合第一方面或上述可能的实施方式，在第一方面的第十三种可能的实施方式中，更新模型参数之前，还包括：
计算模型参数的更新梯度；
更新模型参数，包括：
基于更新梯度、预设更新系数和更新前的模型参数，计算更新后的模型参数。
结合第一方面或上述可能的实施方式，在第一方面的第十四种可能的实施方式中，在计算模型参数的更新梯度之前，还包括：
判断是否执行粗粒度训练控制指令；
计算模型参数的更新梯度，包括：
当判定执行粗粒度训练控制指令时，通过第一预设关系计算更新梯度，第一预设关系包括更新梯度等于第一损失函数对模型参数的偏导值，第一损失函数为Loss1＝|vcosα-vsinα-Q|2，其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预设粗粒度训练控制指令之间匹配的程度；
当判定不执行粗粒度训练控制指令时，通过第二预设关系计算更新梯度，第二预设关系包括更新梯度等于第二损失函数对模型参数的偏导值，第二损失函数为Loss2＝|vsinα-vcosα-Q|2。
第二方面，本发明实施例提供了一种车辆的控制装置，包括：
获取单元，用于在自动驾驶车辆处于自动驾驶状态时，获取自动驾驶车辆的行驶状态信息和自动驾驶车辆的第一行驶环境信息；
计算单元，用于基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车辆的粗粒度控制指令，粗粒度控制指令用于控制自动驾驶车辆的行驶方式，控制指令决策模型为基于自动驾驶车辆在训练状态时的训练行驶状态信息和自动驾驶车辆在训练状态时的训练行驶环境信息训练得到的；
判断单元，用于根据第二行驶环境信息判断是否执行粗粒度控制指令，第二行驶环境信息包括在自动驾驶车辆处于自动驾驶状态时，与粗粒度控制指令对应的行驶环境信息；
确定单元，用于当判定执行粗粒度控制指令时，根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令，细粒度控制指令用于控制自动驾驶车辆的行驶参数，车道信息包括自动驾驶车辆所行驶道路中粗粒度控制指令对应车道的信息；
输出单元，用于输出细粒度控制指令。
本发明实施例中，控制装置通过行驶状态信息、第一行驶环境信息和控制指令决策模型先计算出粗粒度控制指令，然后基于第二行驶环境信息来判定是否可以执行粗粒度控制指令，即在确定控制自动驾驶车辆的行驶参数之前，先计算出粗粒度控制指令计算出自动驾驶车辆的行驶方式，当判定可以执行粗粒度控制指令时，再将粗粒度控制指令转换为细粒度控制指令，从而能够避免在车辆自动控制时输出不必要或错误的控制指令，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故；同时，本发明实施例中控制装置在判定可以执行粗粒度控制指令后，基于自动驾驶车辆的行驶状态信息和粗粒度控制指令对应的车道信息将粗粒度控制指令转换为细粒度控制指令，即充分考虑自动驾驶车辆行驶时的自身状况和车道信息确定出适合控制自动驾驶车辆的细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故，从降低自动驾驶的车辆的安全隐患，进一步提高自动驾驶车辆的安全性和舒适性。
结合第二方面，在第二方面的第一种可能的实施方式中，粗粒度控制指令包括直行；
车道信息包括自动驾驶车辆直行时所行驶车道的预设期望速度，自动驾驶车辆的行驶状态信息包括自动驾驶车辆的当前速度；
确定单元具体用于基于预设期望速度和当前速度，确定细粒度控制指令。
本实施方式中，粗粒度控制指令包括直行时，控制装置不需要对自动驾驶车辆的方向盘转角进行控制，可以只控制自动驾驶车辆的行驶速度，所以根据自动驾驶车辆的当前的速度和自动驾驶车辆在直行车道的预设期望速度来对自动驾驶车辆来确定细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性， 避免自动驾驶车辆出现左右摆动等不符合自然驾驶行为。
结合第二方面或上述可能的实施方式，在第二方面的第二种可能的实施方式中，细粒度控制指令包括自动驾驶车辆的油门大小；
确定单元具体用于：
当当前速度大于预设期望速度时，油门大小等于零；
当当前速度不大于预设期望速度时，基于第一预设公式计算自动驾驶车辆的油门大小，其中，第一预设公式包括：
油门大小＝预设油门控制系数×(预设期望速度-当前速度+预设值)。
本实施方式中，当前速度大于预设期望速度时，说明自动驾驶车辆的行驶速度过大，所以此时控制装置将油门大小确定为零，以便于尽快降低自动驾驶车辆的行驶速度；当前速度不大于预设期望速度时，说明自动驾驶车辆的行驶速度正常，此时控制装置基于预设期望速度和当前速度之间的差值和预设油门控制系数确定油门大小，以实现对油门的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中预设期望速度不大于自动驾驶车辆直行时所行驶车道的最高限速。
结合第二方面或上述可能的实施方式，在第二方面的第三种可能的实施方式中，细粒度控制指令包括自动驾驶车辆的刹车力度；
确定单元具体用于：
当当前速度小于预设期望速度时，刹车力度等于零；
当当前速度不小于预设期望速度时，基于第二预设公式计算自动驾驶车辆的刹车力度，其中，第二预设公式包括：
刹车力度＝预设刹车控制系数×(当前速度-预设期望速度)。
本实施方式中，当前速度小于预设期望速度时，说明自动驾驶车辆的行驶速度正常，此时控制装置可以不进行刹车，所以将刹车力度调整为零；当前速度不小于预设期望速度时，说明自动驾驶车辆的行驶速度过大，此时控制装置需要通过刹车来控制自动驾驶车辆的速度，所以基于预设期望速度和当前速度之间的差值和预设刹车控制系数确定刹车力度，以实现对刹车的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中预设期望速度不大于自动驾驶车辆直行时所行驶车道的最高限速。
结合第二方面或上述可能的实施方式，在第二方面的第四种可能的实施方式中，粗粒度控制指令包括换道方向；
确定单元具体用于：
基于自动驾驶车辆的行驶状态信息，模拟自动驾驶车辆在换道方向的换道行驶路径；
根据换道行驶路径和车道信息，确定自动驾驶车辆的细粒度控制指令。
本实施方式中，粗粒度控制指令包括换道方向时，控制装置控制自动驾驶车辆向换道方向侧进行换道，此时首先模拟出自动驾驶车辆的换道行驶路径，即在自动驾驶车辆实际进行换道之前，先确定出换道路径，然后再基于预先确定的换道路径 和车道信息确定自动驾驶车辆在换道时的细粒度控制指令，从而提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，进而提高自动驾驶车辆换道时的安全性。
结合第二方面或上述可能的实施方式，在第二方面的第五种可能的实施方式中，细粒度控制指令包括方向盘转角；
车道信息包括换道行驶路径对应车道的车道宽度和换道行驶路径对应车道的车道中线；
确定单元具体用于：
确定在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度；
在换道行驶路径中，根据换道行驶路径对应车道的车道宽度确定自动驾驶车辆在转向角度对应位置的所属车道的目标车道宽度；
在换道行驶路径中，根据换道行驶路径对应车道的车道中线确定自动驾驶车辆在转向角度对应位置与所属车道的车道中线之间的目标距离；
基于第三预设公式、转向角度、目标车道宽度和目标距离计算方向盘转角，其中，第三预设公式包括：
方向盘转角＝[转向角度-目标距离/(目标车道宽度×第一转角系数)]×第二转角系数。
本实施方式中，由于自动驾驶车辆的换道行驶路径为曲线，控制装置需要对自动驾驶车辆的方向盘转角进行控制，所以基于预先模拟的换道行驶路径可以确定出自动驾驶车辆在换道时需要转向的位置和转向角度，然后在结合换道行驶路径中各转向角度对应车道的信息确定出方向盘在对应各转向角度时的转角，即确定出自动驾驶车辆在换道时方向盘需要转动的角度，从而实现对方向盘的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中预设期望速度不大于自动驾驶车辆直行时所行驶车道的最高限速。方向盘转角的取值范围可以为大于等于-1且小于等于1。
结合第二方面或上述可能的实施方式，在第二方面的第六种可能的实施方式中，细粒度控制指令还包括自动驾驶车辆的油门大小；
车道信息还包括换道行驶路径对应车道的预设期望速度；
确定单元具体用于：
在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置的当前速度；
在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度；
当当前速度大于目标预设期望速度时，油门大小等于零；
当当前速度不大于目标预设期望速度时，基于第四预设公式计算自动驾驶车辆在转向角度对应位置的油门大小，其中，第四预设公式包括：
油门大小＝预设油门控制系数×(目标预设期望速度-当前速度+预设值)。
本实施方式中，在自动驾驶车辆换道时，控制装置还可以对行驶速度进行控制。在当前速度大于目标预设期望速度时，将油门大小确定为零，以便于尽快降低自动 驾驶车辆的行驶速度；当前速度不大于目标预设期望速度时，基于目标预设期望速度和当前速度之间的差值和预设油门控制系数确定油门大小，以实现对油门的精确控制，同时保证自动驾驶车辆行驶的安全性。
具体的，本实施方式中目标预设期望速度不大于换道行驶路径对应车道的最高限速。
结合第二方面或上述可能的实施方式，在第二方面的第七种可能的实施方式中，细粒度控制指令还包括自动驾驶车辆的刹车力度；
车道信息还包括换道行驶路径对应车道的预设期望速度；
确定单元具体用于：
在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置的当前速度；
在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度；
当当前速度小于目标预设期望速度时，刹车力度等于零；
当当前速度不小于目标预设期望速度时，基于第五预设公式计算自动驾驶车辆在转向角度对应位置的刹车力度，其中，第五预设公式包括：
刹车力度＝预设刹车控制系数×(当前速度-目标预设期望速度)。
本实施方式中，在自动驾驶车辆换道时，控制装置还可以对行驶速度进行控制。当前速度小于目标预设期望速度时，可以不进行刹车，所以将刹车力度调整为零；当前速度不小于目标预设期望速度时，需要通过刹车来控制自动驾驶车辆的速度，所以基于目标预设期望速度和当前速度之间的差值和预设刹车控制系数确定刹车力度，以实现对刹车的精确控制，保证自动驾驶车辆行驶的安全性。
具体的，在上述实施方式中目标预设期望速度不大于换道行驶路径对应车道的最高限速。油门大小的取值范围可以为大于等于0且小于等于1。刹车力度的取值范围可以为大于等于0且小于等于1。预设值的取值可以为1。
结合第二方面或上述可能的实施方式，在第二方面的第八种可能的实施方式中，第一行驶环境信息包括自动驾驶车辆所行驶车道的车道信息，自动驾驶车辆预设距离内车辆的信息，自动驾驶车辆预设距离内路面的信息中至少一项；
第二行驶环境信息包括自动驾驶车辆所行驶车道中与粗粒度控制指令对应的车道信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的路面信息中至少一项。
结合第二方面或上述可能的实施方式，在第二方面的第九种可能的实施方式中，第二行驶环境信息包括目标车辆与自动驾驶车辆之间的第一车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆；
判断单元具体用于：
当第一车辆距离大于安全距离时，判定执行粗粒度控制指令；
当第一车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
本实施方式中，在计算出粗粒度控制指令后，从安全角度出发，控制装置可以通过判断自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆是否会影响自动驾 驶车辆的安全来确定是否执行粗粒度控制指令，即判断目标车辆与自动驾驶车辆之间的第一车辆距离是否达到安全距离。通过对第一车辆距离和安全距离之间大小的判断，在第一车辆距离未达到安全距离时不执行粗粒度控制指令，在第一车辆距离达到安全距离时才执行粗粒度控制指令，可以有效的提高自动驾驶车辆在自动驾驶中的安全性，避免自动驾驶车辆执行错误或不必要的控制指令导致安全事故的发生。
结合第二方面或上述可能的实施方式，在第二方面的第十种可能的实施方式中，粗粒度控制指令包括换道方向；装置还包括：
模拟单元，用于基于自动驾驶车辆的行驶状态信息模拟自动驾驶车辆在换道方向的换道行驶路径；
第二行驶环境信息包括目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆；
判断单元具体用于：
当第二车辆距离大于安全距离时，判定执行粗粒度控制指令；
当第二车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
本实施方式中，粗粒度控制指令包括换道方向时，控制装置可以先模拟出自动驾驶车辆的换道行驶路径，然后基于目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离对是否执行粗粒度控制指令进行判断，从而提高判断的准确性和自动驾驶车辆在自动驾驶中的安全性。
具体的，在上述实施例中，安全距离可以为：自动驾驶车辆以自动驾驶车辆的当前速度行驶、且目标车辆以目标车辆的当前速度行驶，在预设时间段内使自动驾驶车辆和目标车辆之间不接触的距离。
结合第二方面或上述可能的实施方式，在第二方面的第十一种可能的实施方式中，还包括：
初始化单元，用于初始化控制指令决策模型的模型参数；
获取单元还用于获取训练参数，训练参数包括训练行驶状态信息和训练行驶环境信息；
计算单元还用于根据模型参数和训练参数计算自动驾驶车辆的粗粒度训练控制指令；以及，用于依据粗粒度训练控制指令计算损失函数的值；
装置还包括：
更新单元，用于当损失函数的值未达到预设条件时，更新模型参数；
计算单元，还用于根据更新后的模型参数和训练参数计算自动驾驶车辆的更新粗粒度训练控制指令，并重新计算损失函数的值，直到损失函数的值达到预设条件；
确定单元还用于将损失函数的值达到预设条件时对应的模型参数确定为控制指令决策模型的最终模型参数。
结合第二方面或上述可能的实施方式，在第二方面的第十二种可能的实施方式中，损失函数Loss1包括：
Loss1＝|vcosα-vsinα-Q|2，
其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预 设粗粒度训练控制指令之间匹配的程度。
结合第二方面或上述可能的实施方式，在第二方面的第十三种可能的实施方式中，计算单元还用于计算模型参数的更新梯度；
更新单元具体用于基于更新梯度、预设更新系数和更新前的模型参数，计算更新后的模型参数。
结合第二方面或上述可能的实施方式，在第二方面的第十四种可能的实施方式中，判断单元还用于判断是否执行粗粒度训练控制指令；
计算单元具体用于：
当判定执行粗粒度训练控制指令时，通过第一预设关系计算更新梯度，第一预设关系包括更新梯度等于第一损失函数对模型参数的偏导值，第一损失函数为Loss1＝|vcosα-vsinα-Q|2，其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预设粗粒度训练控制指令之间匹配的程度；
当判定不执行粗粒度训练控制指令时，通过第二预设关系计算更新梯度，第二预设关系包括更新梯度等于第二损失函数对模型参数的偏导值，第二损失函数为Loss2＝|vsinα-vcosα-Q|2。
第三方面，本发明实施例提供了一种车辆的控制设备，包括：
存储器、处理器和总线；
存储器和处理器通过总线连接并完成相互间的通信；
存储器用于存储程序代码；
处理器通过读取存储器中存储的可执行程序代码来运行与可执行程序代码对应的程序，以用于执行如第一方面所述的方法。
第四方面，本发明实施例提供一种计算机可读存储介质，计算机可读存储介质中存储有指令，当其在计算机上运行时，使得计算机执行第一方面所述的方法。
第五方面，本发明实施例提供了一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行第一方面所述的方法。
第六方面，本发明实施例提供一种计算机程序，当其在计算机上运行时，使得计算机执行第一方面所述的方法。
附图说明
为了更清楚地说明本发明实施例的技术方案，下面将对本发明实施例中所需要使用的附图作简单地介绍。
图1是根据本发明一实施例提供的车辆的控制方法的示意性流程图；
图2是根据本发明一实施例中一种确定细粒度控制指令的方法的示意性流程图；
图3是根据本发明一实施例中又一种确定细粒度控制指令的方法的示意性流程图；
图4是根据本发明一实施例的车辆的控制装置的示意性框图；
图5是根据本发明又一实施例的车辆的控制装置的示意性框图；
图6是根据本发明一实施例的车辆的控制设备的示意性框图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述。
本发明实施例适用于对车辆进行自动驾驶控制的场景。本发明实施例中对自动驾驶车辆进行控制可以为控制决策系统，控制决策系统可以分为两部分，分别为强化学习层和运动轨迹安全控制层。强化学习层可以通过行驶策略决策模型，再根据获取的自动驾驶车辆相关的信息计算出粗粒度控制指令，然后将粗粒度控制指令传输给运动轨迹安全控制层；运动轨迹安全控制层可以从行驶安全角度对是否执行粗粒度控制指令进行判断，当判定执行粗粒度控制指令时，将粗粒度控制指令转换为细粒度控制指令，即控制自动驾驶车辆行驶的具体行驶参数，并输出细粒度控制指令至车辆的控制系统，以实现对车辆的自动驾驶控制。
图1示出了根据本发明一实施例的车辆的控制方法的示意性流程图。如图1所示，该方法包括以下步骤101-105。
101，在自动驾驶车辆处于自动驾驶状态时，获取自动驾驶车辆的行驶状态信息和自动驾驶车辆的第一行驶环境信息。
其中，在自动驾驶车辆处于自动驾驶状态时，可以实时获取自动驾驶车辆的行驶状态信息和第一行驶环境信息，然后基于获取的行驶状态信息和第一行驶环境信息来计算自动驾驶车辆的粗粒度控制指令。
本发明实施例中，第一行驶环境信息可以包括自动驾驶车辆所行驶车道的车道信息，自动驾驶车辆预设距离内车辆的信息，自动驾驶车辆预设距离内路面的信息中至少一项。车道信息可以包括车道的最高限速和车道的宽度等等，车辆的信息可以包括车辆数量、车辆行驶方向以及车辆与自动驾驶车辆之间的距离等等，路面的信息可以包括路面上隔离设施的信息和路面上障碍物的信息等等。自动驾驶车辆的行驶状态信息可以包括自动驾驶车辆的位置、自动驾驶车辆的速度、自动驾驶车辆的行驶方向和自动驾驶车辆与其所行驶车道的夹角等等。
需要说明的是，本步骤中对获取行驶状态信息和第一行驶环境信息的方式不做限定，可以包括各类传感器，如激光雷达、超声波雷达、毫米波雷达等，车载摄像头，全球定位系统(Global Positioning System，GPS)，地图，自动驾驶车辆的车载诊断系统(OBD)数据等等。
102，基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车辆的粗粒度控制指令。
其中，粗粒度控制指令用于控制自动驾驶车辆的行驶方式，控制指令决策模型为基于自动驾驶车辆在训练状态时的训练行驶状态信息和自动驾驶车辆在训练状态时的训练行驶环境信息训练得到的。
本发明实施例中，行驶方式可以包括直行、换道、掉头、转弯等等，粗粒度控制指令可以包括直行或换道，在粗粒度控制指令包括换道时，还可以包括换道方向。控制指令决策模型为预先训练得出的，其在训练时将训练行驶状态信息和训练行驶环境信息作为输入。
103，根据第二行驶环境信息判断是否执行粗粒度控制指令。
其中，第二行驶环境信息包括在自动驾驶车辆处于自动驾驶状态时，与粗粒度控制指令对应的行驶环境信息。本步骤中判断步骤102计算的粗粒度控制指令是否可以执行主要参考自动驾驶车辆行驶时的安全因素，来避免自动驾驶车辆执行步骤102计算的错误或必要的粗粒度控制指令。
由于步骤102计算出粗粒度控制指令后，已经确定自动驾驶车辆的行驶方式，所以在本步骤判断是否执行粗粒度控制指令时，只需要根据对粗粒度控制指令执行有影响的行驶环境信息进行判断即可，所以本步骤中基于与粗粒度控制指令对应的行驶环境信息来判断是否执行粗粒度控制指令。
具体的，当粗粒度控制指令包括直行时，与粗粒度控制指令对应的行驶环境信息可以包括自动驾驶车辆前方预设范围内的行驶环境信息；当粗粒度控制指令包括换道方向时，与粗粒度控制指令对应的行驶环境信息可以包括自动驾驶车辆在换道方向一侧预设范围内的行驶环境信息。
本发明实施例中，第二行驶环境信息可以包括自动驾驶车辆所行驶车道中与粗粒度控制指令对应的车道信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的路面信息中至少一项。车道信息可以包括车道的最高限速和车道的宽度等等，车辆的信息可以包括车辆数量、车辆行驶方向和车辆与自动驾驶车辆之间的距离等等，路面的信息可以包括路面上隔离设施的信息和路面上障碍物的信息等等。
需要说明的是，本步骤中第二行驶环境信息可以实时进行获取，也可以从步骤101获取的第一行驶环境信息中确定。
104，当判定执行粗粒度控制指令时，根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令。
其中，细粒度控制指令可以为用于控制自动驾驶车辆的行驶参数，可以包括方向盘转角、油门大小和刹车力度等等。车道信息可以包括自动驾驶车辆所行驶道路中粗粒度控制指令对应车道的信息，自动驾驶车辆所行驶道路包括自动驾驶车辆所行驶的车道和自动驾驶车辆允许换道的车道，自动驾驶车辆所行驶道路中粗粒度控制指令对应车道可以表示自动驾驶车辆在执行粗粒度控制指令所要行驶的车道。例如，粗粒度控制指令包括直行时，自动驾驶车辆所行驶道路中粗粒度控制指令对应车道可以为自动驾驶车辆直行时所行驶车道；粗粒度控制指令包括换道方向时，自动驾驶车辆所行驶道路中粗粒度控制指令对应车道可以为自动驾驶车辆在换道时所行驶的车道。
需要说明的是，本步骤中车道信息和自动驾驶车辆的行驶状态信息可以实时进行获取，也可以从步骤101获取的行驶状态信息和第一行驶环境信息中确定。
105，输出细粒度控制指令。
其中，本步骤中将步骤104确定的细粒度控制指令输出，实现对自动驾驶车辆相应的部件进行控制，从而达到车辆自动驾驶的目的。
本发明实施例中，通过行驶状态信息、第一行驶环境信息和控制指令决策模型先计算出粗粒度控制指令，然后基于第二行驶环境信息来判定是否可以执行粗粒度 控制指令，即在确定控制自动驾驶车辆的行驶参数之前，先计算出粗粒度控制指令计算出自动驾驶车辆的行驶方式，当判定可以执行粗粒度控制指令时，再将粗粒度控制指令转换为细粒度控制指令，从而能够避免在车辆自动控制时输出不必要或错误的控制指令，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故；同时，本发明实施例中在判定可以执行粗粒度控制指令后，基于自动驾驶车辆的行驶状态信息和粗粒度控制指令对应的车道信息将粗粒度控制指令转换为细粒度控制指令，即充分考虑自动驾驶车辆行驶时的自身状况和车道信息确定出适合控制自动驾驶车辆的细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故，从降低自动驾驶的车辆的安全隐患，进一步提高自动驾驶车辆的安全性和舒适性。
图2是根据本发明一实施例中一种确定细粒度控制指令的方法的示意性流程图。
在本发明实施例的一种实施方式中，步骤102计算出的粗粒度控制指令可以包括直行，此时步骤104中车道信息包括自动驾驶车辆直行时所行驶车道的预设期望速度，自动驾驶车辆的行驶状态信息包括自动驾驶车辆的当前速度；则当判定执行粗粒度控制指令时，如图2所示，步骤104可以具体执行为步骤1041。
1041，基于预设期望速度和当前速度，确定细粒度控制指令。
其中，由于粗粒度控制指令为直行，所以不需要对自动驾驶车辆的方向盘转角进行控制，方向盘可以保持当前状态不进行调整，此时可以控制自动驾驶车辆的行驶速度。
本发明实施例中，自动驾驶车辆在行驶时可以针对不同的行驶车道预设对应的期望速度，为了符合车辆行驶规则和保证行驶安全，预设期望速度不能大于车道的最高限速。本步骤中基于自动驾驶车辆的当前速度和其直行时所行驶车道的预设期望速度来对自动驾驶车辆的速度进行控制，即计算细粒度控制指令，使其能够安全行驶。
对车辆行驶速度进行控制通常通过控制车辆的油门或刹车来实现。
具体的，本发明实施例中，细粒度控制指令可以包括自动驾驶车辆的油门大小。
在确定油门大小时，需要首先判定自动驾驶车辆的当前速度和其直行时所行驶车道的预设期望速度之间的大小关系。在当前速度大于预设期望速度时，说明自动驾驶车辆的行驶速度过大，所以此时油门可以设置为最小状态，将油门大小确定为零，以便于尽快降低自动驾驶车辆的行驶速度；在当前速度不大于预设期望速度时，说明自动驾驶车辆的行驶速度正常，此时基于预设期望速度和当前速度之间的差值和预设油门控制系数确定油门大小以实现对油门的精确控制，保证自动驾驶车辆行驶的安全性。在当前速度不大于预设期望速度时，可以基于第一预设公式计算自动驾驶车辆的油门大小，其中，第一预设公式如公式1所示。
油门大小＝预设油门控制系数×(预设期望速度-当前速度+预设值)    (1)
其中，在公式1中，预设期望速度表示自动驾驶车辆直行时所行驶车道的预设期望速度，当自动驾驶车辆直行时所行驶车道的预设期望速度为自动驾驶车辆直行时所行驶车道的最高限速时，公式1中预设期望速度可以替换为最高限速的值。公式1中当前速度表示自动驾驶车辆的当前速度。
需要说明的是，通过上述公式1可以计算出自动驾驶车辆油门大小，通常情况下，设置油门大小为1时表示油门最大状态，油门大小为0时表示油门最小的状态，所以油门大小的取值范围可以为大于等于0且小于等于1。预设值可以根据实际应用场景和车辆的具体性能来取值，通常情况下，预设值可以等于1。
具体的，本发明实施例中，细粒度控制指令可以包括自动驾驶车辆的刹车力度。
在确定刹车力度时，需要首先判定自动驾驶车辆的当前速度和其直行时所行驶车道的预设期望速度之间的大小关系。当前速度小于预设期望速度时，说明自动驾驶车辆的行驶速度正常，此时自动驾驶车辆可以不进行刹车，所以刹车可以设置为最小状态，即刹车力度确定为零；当前速度不小于预设期望速度时，说明自动驾驶车辆的行驶速度过大，此时需要通过刹车来降低自动驾驶车辆的速度，所以基于预设期望速度和当前速度之间的差值和预设刹车控制系数确定刹车力度，以实现对刹车的精确控制，保证自动驾驶车辆行驶的安全性。在当前速度不小于预设期望速度时，可以基于第二预设公式计算自动驾驶车辆的刹车力度，其中，第二预设公式如公式2所示。
刹车力度＝预设刹车控制系数×(当前速度-预设期望速度)      (2)
其中，在公式2中，预设期望速度表示自动驾驶车辆直行时所行驶车道的预设期望速度，当自动驾驶车辆直行时所行驶车道的预设期望速度为自动驾驶车辆直行时所行驶车道的最高限速时，公式2中预设期望速度可以替换为最高限速的值。公式2中当前速度表示自动驾驶车辆的当前速度。
需要说明的是，通过上述公式2可以计算出自动驾驶车辆的刹车力度，通常情况下，设置刹车力度为1时表示刹车力度的最大状态，刹车力度为0时表示刹车力度最小的状态，所以刹车力度的取值范围可以为大于等于0且小于等于1的数值。
需要说明的是，公式1中预设油门控制系数和公式2中预设刹车控制系数可以根据自动驾驶车辆的性能进行确定，不同品牌和型号的车辆，其对应的预设油门控制系数和预设刹车控制系数均可以不同，例如，预设刹车控制系数可以为0.1，预设油门控制系数可以为0.2。
本发明实施例中，粗粒度控制指令包括直行时，不需要对自动驾驶车辆的方向盘转角进行控制，可以只控制自动驾驶车辆的行驶速度，所以根据自动驾驶车辆的当前的速度和自动驾驶车辆在直行车道的预设期望速度来对自动驾驶车辆来确定细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，避免自动驾驶车辆出现左右摆动等不符合自然驾驶行为。
图3是根据本发明一实施例中又一种确定细粒度控制指令的方法的示意性流程图。
在本发明实施例的又一种实施方式中，步骤102计算出的粗粒度控制指令可以包括换道方向，此时当判定执行粗粒度控制指令时，如图3所示，步骤104可以具体执行为步骤1042和步骤1043。
1042，基于自动驾驶车辆的行驶状态信息，模拟自动驾驶车辆在换道方向的换道行驶路径。
其中，模拟换道行驶路径的实现方式不做限定。
1043，根据换道行驶路径和车道信息，确定自动驾驶车辆的细粒度控制指令。
本发明实施例中，在粗粒度控制指令包括换道方向时，说明自动驾驶车辆需要在换道方向对应一侧进行换道，此时首先模拟出自动驾驶车辆的换道行驶路径，即在自动驾驶车辆实际进行换道之前，先确定出换道路径，然后再基于预先确定的换道路径和车道信息确定自动驾驶车辆在换道时的细粒度控制指令，从而提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，进而提高自动驾驶车辆换道时的安全性。
具体的，本发明实施例中，细粒度控制指令可以包括自动驾驶车辆的方向盘转角。此时步骤1043中车道信息包括换道行驶路径对应车道的车道宽度和换道行驶路径对应车道的车道中线；步骤1043可以具体执行为：确定在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度；在换道行驶路径中，根据换道行驶路径对应车道的车道宽度确定自动驾驶车辆在转向角度对应位置的所属车道的目标车道宽度；在换道行驶路径中，根据换道行驶路径对应车道的车道中线确定自动驾驶车辆在转向角度对应位置与所属车道的车道中线之间的目标距离；基于第三预设公式、转向角度、目标车道宽度和目标距离计算方向盘转角，其中，第三预设公式如公式3所示。
其中，模拟出的换道行驶路径为一段自动驾驶车辆换道行驶的曲线路径，本发明实施例中，为了便于对自动驾驶车辆的控制，可以将换道行驶路径中自动驾驶车辆的连续行驶过程离散为至少一个时间点，并计算自动驾驶车辆在这些时间点的细粒度控制指令，以此种方式来实现对自动驾驶车辆整个换道行驶过程的控制。所以针对将换道行驶路径离散后的时间点可以确定出自动驾驶车辆在换道行驶路径的至少一个位置，进而可以模拟出在换道行驶中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间的至少一个转向角度，即确定出在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度。
在确定出各转向角度后，需要将这些转向角度转换为自动驾驶车辆的方向盘转角。针对至少一个转向角度中每个转向角度，可以确定出此转向角度对应换道行驶路径中的位置，即对应此转向角度时自动驾驶车辆在换道行驶路径中的位置，进而确定出自动驾驶车辆在换道行驶路径中位置所属的车道，以及自动驾驶车辆在换道行驶路径中位置所属车道的目标车道宽度。同样的，针对至少一个转向角度中每个转向角度，可以根据换道行驶路径对应车道的车道中线，确定出自动驾驶车辆在换道行驶路径中转向角度对应位置所属车道的车道中线与自动驾驶车辆在换道行驶路径中转向角度对应位置之间的目标距离。在针对至少一个转向角度中每个转向角度，确定出对应的目标车道宽度和目标距离后，再根据公式3即可进行计算出转向角度对应的方向盘转角。
方向盘转角＝[转向角度-目标距离/(目标车道宽度×第一转角系数)]×第二转角系数
                                                     (3)
其中，在公式3中，转向角度表示自动驾驶车辆行驶方向与自动驾驶车辆直行方向之间的角度，目标距离表示自动驾驶车辆在换道行驶路径中转向角度对应位置与自动驾驶车辆在换道行驶路径中转向角度对应位置所属车道的车道中线之间的距 离，目标车道宽度表示自动驾驶车辆在换道行驶路径中转向角度对应位置的所属车道的目标车道宽度。
需要说明的是，在公式3中第一转角系数和第二转角系数可以根据自动驾驶车辆的性能进行确定，不同品牌和型号的车辆，其对应的第一转角系数和第二转角系数均可以不同，例如，第一转角系数可以取值为0.541，第一转角系数可以取值为0.4。本发明实施例中，可以设置方向盘在向一侧转动最大角度时方向盘转角的取值为-1，方向盘在向另一侧转动最大角度时方向盘转角的取值为1，则方向盘转角的范围可以为大于等于-1且小于等于1。
本发明实施例中，由于自动驾驶车辆的换道行驶路径为曲线，需要对自动驾驶车辆的方向盘转角进行控制，所以基于预先模拟的换道行驶路径可以确定出自动驾驶车辆在换道时需要转向的位置和转向角度，然后在结合换道行驶路径中各转向角度对应车道的信息确定出方向盘在对应各转向角度时的转角，即确定出自动驾驶车辆在换道时方向盘需要转动的角度，从而实现对方向盘的精确控制，保证自动驾驶车辆行驶的安全性。
本发明实施例中，在确定出自动驾驶车辆换道过程中方向盘转角后，还可以对自动驾驶车辆的行驶速度进行控制。在上述步骤1043执行过程中，为了便于对自动驾驶车辆的控制，将换道行驶路径中自动驾驶车辆的连续行驶过程离散为至少一个时间点，并确定出至少一个转向角度。本发明实施例中，可以通过确定自动驾驶车辆在换道行驶路径中对应至少一个转向角度中每个转向角度的速度，对进行自动驾驶车辆的行驶速度进行控制。
具体的，细粒度控制指令还包括自动驾驶车辆的油门大小。此时步骤104中车道信息还可以包括换道行驶路径对应车道的预设期望速度。
上述步骤1043的执行过程中，在确定出至少一个转向角度之后，还可以执行如下过程：在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置的当前速度；在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度；当当前速度大于目标预设期望速度时，油门大小等于零；当当前速度不大于目标预设期望速度时，基于第四预设公式计算自动驾驶车辆在转向角度对应位置的油门大小。
其中，本发明实施例中，对至少一个转向角度中每个转向角度均可以确定对应的油门大小。首先确定出在换道行驶路径中，自动驾驶车辆在转向角度对应位置的当前速度和自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度，根据公式4进行计算即可确定油门大小。
油门大小＝预设油门控制系数×(目标预设期望速度-当前速度+预设值)    (4)
其中，公式4中目标预设期望速度表示自动驾驶车辆在换道行驶路径中转向角度对应位置所属车道的目标预设期望速度，当目标预设期望速度为自动驾驶车辆在换道行驶路径中转向角度对应位置所属车道的最高限速时，公式4中目标预设期望速度可以替换为最高限速的值。公式4中当前速度表示自动驾驶车辆在换道行驶路径中转向角度对应位置的当前速度。
本发明实施例中，在自动驾驶车辆换道时，还可以对行驶速度进行控制。在当前速度大于目标预设期望速度时，将油门大小确定为零，以便于尽快降低自动驾驶车辆的行驶速度；当前速度不大于目标预设期望速度时，基于目标预设期望速度和当前速度之间的差值和预设油门控制系数确定油门大小，以实现对油门的精确控制，同时保证自动驾驶车辆行驶的安全性。
具体的，细粒度控制指令还可以包括自动驾驶车辆的刹车力度。此时步骤104中车道信息还可以包括换道行驶路径对应车道的预设期望速度。
在上述步骤1043的执行过程中，在确定出至少一个转向角度之后，还可以执行如下过程：在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置的当前速度；在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度；当当前速度小于目标预设期望速度时，刹车力度等于零；当当前速度不小于目标预设期望速度时，基于第五预设公式计算自动驾驶车辆在转向角度对应位置的刹车力度。
其中，本发明实施例中，对至少一个转向角度中每个转向角度均可以确定对应的刹车力度。首先确定出在换道行驶路径中，自动驾驶车辆在转向角度对应位置的当前速度和自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度，根据公式5进行计算即可确定刹车力度。
刹车力度＝预设刹车控制系数×(当前速度-目标预设期望速度)    (5)
其中，公式5中目标预设期望速度表示自动驾驶车辆在换道行驶路径中转向角度对应位置所属车道的目标预设期望速度，当目标预设期望速度为自动驾驶车辆在换道行驶路径中转向角度对应位置所属车道的最高限速时，公式5中目标预设期望速度可以替换为最高限速的值。公式5中当前速度表示自动驾驶车辆在换道行驶路径中转向角度对应位置的当前速度。
本发明实施例中，在自动驾驶车辆换道时，还可以对行驶速度进行控制。当前速度小于目标预设期望速度时，可以不进行刹车，所以将刹车力度调整为零；当前速度不小于目标预设期望速度时，需要通过刹车来控制自动驾驶车辆的速度，所以基于目标预设期望速度和当前速度之间的差值和预设刹车控制系数确定刹车力度，以实现对刹车的精确控制，保证自动驾驶车辆行驶的安全性。
需要说明的是，公式4中油门大小、预设油门控制系数和预设值的取值方式与公式1中油门大小、预设油门控制系数和预设值的取值方式相同。公式5中刹车力度和预设刹车控制系数的取值方式与公式2中油门大小和预设刹车控制系数的取值方式相同。
作为本发明实施例中又一种可选的实施方式，第二行驶环境信息可以包括目标车辆与自动驾驶车辆之间的第一车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆。此时步骤103可以具体执行为：当第一车辆距离大于安全距离时，判定执行粗粒度控制指令；当第一车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
其中，当粗粒度控制指令包括直行时，目标车辆表示自动驾驶车辆前方预设距 离内的车辆；当粗粒度控制指令包括换道方向时，目标车辆表示自动驾驶车辆在换道方向一侧预设距离内的车辆。如果自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆为零，即自动驾驶车辆预设距离内不存在与粗粒度控制指令对应的车辆，则可以将第一车辆距离确定为无限大。
需要说明的是，当第一车辆距离小于或等于预设安全距离时，可以判定不执行粗粒度控制指令，此时自动驾驶车辆可以丢弃粗粒度控制指令，保持当前的驾驶状态。在粗粒度控制指令包括直行时，如果判定不执行粗粒度控制指令，也可以执行本发明实施例中步骤1041中所述的处理过程，以便于对自动驾驶车辆实时进行行驶速度的控制。在本发明实施例中，还可以通过其他信息判断是否执行粗粒度控制指令，本发明实施例不做限定。
本发明实施例中，安全距离可以为：自动驾驶车辆以自动驾驶车辆的当前速度行驶、且目标车辆以目标车辆的当前速度行驶，在预设时间段内使自动驾驶车辆和目标车辆之间不接触的距离。即安全距离需要保证自动驾驶车辆以自动驾驶车辆的当前速度行驶预设时间段，同时目标车辆以目标车辆的当前速度行驶相同的预设时间段，在此行驶过程中自动驾驶车辆和目标车辆不会发生碰撞。
本发明实施例中，在计算出粗粒度控制指令后，从安全角度出发，可以通过判断自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆是否会影响自动驾驶车辆的安全来确定是否执行粗粒度控制指令，即判断目标车辆与自动驾驶车辆之间的第一车辆距离是否达到安全距离。通过对第一车辆距离和安全距离之间大小的判断，在第一车辆距离未达到安全距离时不执行粗粒度控制指令，在第一车辆距离达到安全距离时才执行粗粒度控制指令，可以有效的提高自动驾驶车辆在自动驾驶中的安全性，避免自动驾驶车辆执行错误或不必要的控制指令导致安全事故的发生。
作为本发明实施例中又一种可选的实施方式，当粗粒度控制指令包括换道方向时，在步骤103之前还可以执行步骤106：基于自动驾驶车辆的行驶状态信息模拟自动驾驶车辆在换道方向的换道行驶路径。此时第二行驶环境信息包括目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆。则步骤103可以具体执行为当第二车辆距离大于安全距离时，判定执行粗粒度控制指令；当第二车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
其中，当粗粒度控制指令包括换道方向时，目标车辆表示自动驾驶车辆在换道方向一侧预设距离内的车辆。如果自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆为零，即自动驾驶车辆预设距离内不存在与粗粒度控制指令对应的车辆，则可以将第二车辆距离确定为无限大。
需要说明的是，当第二车辆距离小于或等于预设安全距离时，可以判定不执行粗粒度控制指令，此时自动驾驶车辆可以丢弃粗粒度控制指令，保持当前的驾驶状态。在本发明实施例中，还可以通过其他信息判断是否执行粗粒度控制指令，本发明实施例不做限定。
本发明实施例中，安全距离可以为：自动驾驶车辆以自动驾驶车辆的当前速度行驶、且目标车辆以目标车辆的当前速度行驶，在预设时间段内使自动驾驶车辆和 目标车辆之间不接触的距离。即安全距离需要保证自动驾驶车辆以自动驾驶车辆的当前速度行驶预设时间段，同时目标车辆以目标车辆的当前速度行驶相同的预设时间段，在此行驶过程中自动驾驶车辆和目标车辆不会发生碰撞。
本发明实施例中，粗粒度控制指令包括换道方向时，可以先模拟出自动驾驶车辆的换道行驶路径，然后基于目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离对是否执行粗粒度控制指令进行判断，从而提高判断的准确性和自动驾驶车辆在自动驾驶中的安全性。
作为本发明实施例中又一种可选的实施方式，在执行步骤102之前，本发明实施例还包括对控制指令决策模型进行训练的过程。具体可以包括如下步骤。
A、初始化控制指令决策模型的模型参数。
其中，在开始对控制指令决策模型进行训练时，需要对控制指令决策模型的模型参数进行初始化。初始化方式可以为随机生成。
B、获取训练参数。
其中，训练参数包括训练行驶状态信息和训练行驶环境信息，即为自动驾驶车辆在训练状态时的训练行驶状态信息和所述自动驾驶车辆在训练状态时的训练行驶环境信息。训练行驶状态信息与步骤101中行驶状态信息包括的内容可以相同，不同之处在于，训练行驶状态信息为自动驾驶车辆在训练状态时获取的，行驶状态信息为自动驾驶车辆在自动驾驶状态时获取的。训练行驶环境信息与步骤101中行驶环境信息包括的内容可以相同，不同之处在于，训练行驶环境信息为自动驾驶车辆在训练状态时获取的，行驶环境信息为自动驾驶车辆在自动驾驶状态时获取的。
C、根据模型参数和训练参数计算自动驾驶车辆的粗粒度训练控制指令。
其中，此时模型参数在训练过程中需要不断更新，直到达到控制指令决策模型的精度或准确度需求，模型参数的更新方式通过计算出粗粒度训练控制指令是否满足要求来确定的。
D、依据粗粒度训练控制指令计算损失函数的值。
其中，损失函数Loss1可以如公式6所示。
Loss1＝|vcosα-vsinα-Q|2         (6)
其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预设粗粒度训练控制指令之间匹配的程度。
需要说明的是，v表示自动驾驶车辆在训练状态下的行驶速度。
E、当损失函数的值未达到预设条件时，更新模型参数。
其中，本发明实施例中通过计算的损失函数的值与预设条件比较来确定控制指令决策模型中模型参数是否已经满足需求，即控制指令决策模型是否训练完成。当损失函数的值未达到预设条件时，表示模型参数还未满足需求，所以需要更新模型参数，继续对控制指令决策模型进行训练。
需要说明的是，预设条件可以根据模型参数的需求，或控制指令决策模型需要达到的标准来确定。本发明实施例中，预设条件可以包括损失函数的值连续处于预设范围内的次数达到预设门限。
F、根据更新后的模型参数和训练参数计算自动驾驶车辆的更新粗粒度训练控制指令，并重新计算损失函数的值，直到损失函数的值达到预设条件。
其中，在更新模型参数后，可以根据更新模型参数重新执行步骤C，得出更新粗粒度训练控制指令，进而执行步骤D计算出新的损失函数值，再将新的损失函数值与预设条件进行比较，当新的损失函数的值未达到预设条件时，再执行步骤E、F，即如此循环执行，直到步骤D计算出的损失函数的值达到预设条件。
G、将损失函数的值达到预设条件时对应的模型参数，确定为控制指令决策模型的最终模型参数。
其中，本步骤中判定计算出的损失函数的值已经达到预设条件，表示模型参数能够满足需求，控制指令决策模型训练完成，此时损失函数的值达到预设条件时对应的模型参数即为控制指令决策模型训练完成后得出的模型参数。
需要说明的是，控制指令决策模型可以为神经网络模型，其训练控制指令决策模型的算法可以为演员-评论家强化学习算法。行驶策略决策模型为演员网络，而Q的值通过评论家网络来计算。评论家网络也可以为神经网络模型，输入为步骤B获取的训练参数和步骤C计算的粗粒度训练控制指令，输出为Q值，评论家网络的参数在步骤A中进行初始化，在每次步骤E更新模型参数后对评论家网络中参数进行更新，更新的方式不做限定，可以采用反向传播算法实现。
在本发明实施例的一种可选的实施方式中，在步骤E更新模型参数之前，还需要计算模型参数的更新梯度。此时在计算出更新梯度后，可以基于更新梯度、预设更新系数和更新前的模型参数，计算更新后的模型参数。
具体的，可以根据公式7计算更新后的模型参数。
Figure PCTCN2017091095-appb-000001
其中，θ表示更新后的模型参数，θ1表示更新前的模型参数，β表示预设更新系数，
Figure PCTCN2017091095-appb-000002
表示更新梯度。
具体的，在上述计算模型参数的更新梯度的过程中，在计算模型参数的更新梯度之前，还可以判断自动驾驶车辆是否会执行粗粒度训练控制指令；此时计算模型参数的更新梯度可以具体执行为如下过程：当判定执行粗粒度训练控制指令时，通过第一预设关系计算更新梯度；当判定不执行训练驾驶策略时，通过第二预设关系计算更新梯度。
其中，第一预设关系可以包括：更新梯度等于第一损失函数对模型参数的偏导值，此时第一损失函数如公式6所示。第二预设关系可以包括：更新梯度等于第二损失函数对模型参数的偏导值，第二损失函数如公式8所示。
Loss2＝|-R-Q|2＝|vsinα-vcosα-Q|2        (8)
公式8中各参数的含义与公式6中个参数的含义相同，在此不再赘述。
当判定执行粗粒度训练控制指令时，公式7中
Figure PCTCN2017091095-appb-000003
当判定不执行粗粒度训练控制指令时，公式7中
Figure PCTCN2017091095-appb-000004
图4是根据本发明一实施例的车辆的控制装置200的示意性框图。如图4所示，该装置200包括：
获取单元201，用于在自动驾驶车辆处于自动驾驶状态时，获取自动驾驶车辆的行驶状态信息和自动驾驶车辆的第一行驶环境信息；
计算单元202，用于基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车辆的粗粒度控制指令，粗粒度控制指令用于控制自动驾驶车辆的行驶方式，控制指令决策模型为基于自动驾驶车辆在训练状态时的训练行驶状态信息和自动驾驶车辆在训练状态时的训练行驶环境信息训练得到的；
判断单元203，用于根据第二行驶环境信息判断是否执行粗粒度控制指令，第二行驶环境信息包括在自动驾驶车辆处于自动驾驶状态时，与粗粒度控制指令对应的行驶环境信息；
确定单元204，用于当判定执行粗粒度控制指令时，根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令，细粒度控制指令用于控制自动驾驶车辆的行驶参数，车道信息包括自动驾驶车辆所行驶道路中粗粒度控制指令对应车道的信息；
输出单元205，用于输出细粒度控制指令。
本发明实施例中，控制装置200通过行驶状态信息、第一行驶环境信息和控制指令决策模型先计算出粗粒度控制指令，然后基于第二行驶环境信息来判定是否可以执行粗粒度控制指令，即在确定控制自动驾驶车辆的行驶参数之前，先计算出粗粒度控制指令计算出自动驾驶车辆的行驶方式，当判定可以执行粗粒度控制指令时，再将粗粒度控制指令转换为细粒度控制指令，从而能够避免在车辆自动控制时输出不必要或错误的控制指令，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故；同时，本发明实施例中控制装置在判定可以执行粗粒度控制指令后，基于自动驾驶车辆的行驶状态信息和粗粒度控制指令对应的车道信息将粗粒度控制指令转换为细粒度控制指令，即充分考虑自动驾驶车辆行驶时的自身状况和车道信息确定出适合控制自动驾驶车辆的细粒度控制指令，提高细粒度控制指令的准确性和自动驾驶车辆控制的精确性，避免自动驾驶车辆出现不符合自然驾驶行为或发送安全事故，从降低自动驾驶的车辆的安全隐患，进一步提高自动驾驶车辆的安全性和舒适性。
可以理解的是，粗粒度控制指令包括直行；
车道信息包括自动驾驶车辆直行时所行驶车道的预设期望速度，自动驾驶车辆的行驶状态信息包括自动驾驶车辆的当前速度；
确定单元204具体用于基于预设期望速度和当前速度，确定细粒度控制指令。
可以理解的是，细粒度控制指令包括自动驾驶车辆的油门大小；
确定单元204具体用于：
当当前速度大于预设期望速度时，油门大小等于零；
当当前速度不大于预设期望速度时，基于第一预设公式计算自动驾驶车辆的油门大小，其中，第一预设公式包括：
油门大小＝预设油门控制系数×(预设期望速度-当前速度+预设值)。
可以理解的是，细粒度控制指令包括自动驾驶车辆的刹车力度；
确定单元204具体用于：
当当前速度小于预设期望速度时，刹车力度等于零；
当当前速度不小于预设期望速度时，基于第二预设公式计算自动驾驶车辆的刹车力度，其中，第二预设公式包括：
刹车力度＝预设刹车控制系数×(当前速度-预设期望速度)。
可以理解的是，粗粒度控制指令包括换道方向；
确定单元204具体用于：
基于自动驾驶车辆的行驶状态信息，模拟自动驾驶车辆在换道方向的换道行驶路径；
根据换道行驶路径和车道信息，确定自动驾驶车辆的细粒度控制指令。
可以理解的是，细粒度控制指令包括方向盘转角；
车道信息包括换道行驶路径对应车道的车道宽度和换道行驶路径对应车道的车道中线；
确定单元204具体用于：
确定在换道行驶路径中自动驾驶车辆行驶方向与自动驾驶车辆当前所在车道直行方向之间至少一个转向角度；
在换道行驶路径中，根据换道行驶路径对应车道的车道宽度确定自动驾驶车辆在转向角度对应位置的所属车道的目标车道宽度；
在换道行驶路径中，根据换道行驶路径对应车道的车道中线确定自动驾驶车辆在转向角度对应位置与所属车道的车道中线之间的目标距离；
基于第三预设公式、转向角度、目标车道宽度和目标距离计算方向盘转角，其中，第三预设公式包括：
方向盘转角＝[转向角度-目标距离/(目标车道宽度×第一转角系数)]×第二转角系数。
可以理解的是，细粒度控制指令还包括自动驾驶车辆的油门大小；
车道信息还包括换道行驶路径对应车道的预设期望速度；
确定单元204具体用于：
在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置所属车道的当前速度；
在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置的目标预设期望速度；
当当前速度大于目标预设期望速度时，油门大小等于零；
当当前速度不大于目标预设期望速度时，基于第四预设公式计算自动驾驶车辆 在转向角度对应位置的油门大小，其中，第四预设公式包括：
油门大小＝预设油门控制系数×(目标预设期望速度-当前速度+预设值)。
可以理解的是，细粒度控制指令还包括自动驾驶车辆的刹车力度；
车道信息还包括换道行驶路径对应车道的预设期望速度；
确定单元204具体用于：
在换道行驶路径中，基于自动驾驶车辆的行驶状态信息确定自动驾驶车辆在转向角度对应位置的当前速度；
在换道行驶路径中，根据换道行驶路径对应车道的预设期望速度确定自动驾驶车辆在转向角度对应位置所属车道的目标预设期望速度；
当当前速度小于目标预设期望速度时，刹车力度等于零；
当当前速度不小于目标预设期望速度时，基于第五预设公式计算自动驾驶车辆在转向角度对应位置的刹车力度，其中，第五预设公式包括：
刹车力度＝预设刹车控制系数×(当前速度-目标预设期望速度)。
可以理解的是，第一行驶环境信息包括自动驾驶车辆所行驶车道的车道信息，自动驾驶车辆预设距离内车辆的信息，自动驾驶车辆预设距离内路面的信息中至少一项；
第二行驶环境信息包括自动驾驶车辆所行驶车道中与粗粒度控制指令对应的车道信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆信息，自动驾驶车辆预设距离内与粗粒度控制指令对应的路面信息中至少一项。
可以理解的是，第二行驶环境信息包括目标车辆与自动驾驶车辆之间的第一车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆；
判断单元203具体用于：
当第一车辆距离大于安全距离时，判定执行粗粒度控制指令；
当第一车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
图5是根据本发明又一实施例的车辆的控制装置200的示意性框图。
可以理解的是，粗粒度控制指令包括换道方向；
如图5所示，控制装置200还可以包括：
模拟单元206，用于基于自动驾驶车辆的行驶状态信息模拟自动驾驶车辆在换道方向的换道行驶路径；
第二行驶环境信息包括目标车辆与自动驾驶车辆在换道行驶路径行驶时的第二车辆距离，目标车辆表示自动驾驶车辆预设距离内与粗粒度控制指令对应的车辆；
判断单元203具体用于：
当第二车辆距离大于安全距离时，判定执行粗粒度控制指令；
当第二车辆距离不大于安全距离时，判定不执行粗粒度控制指令。
可以理解的是，如图5所示，控制装置200还可以包括：
初始化单元207，用于初始化控制指令决策模型的模型参数；
获取单元201还用于获取训练参数，训练参数包括训练行驶状态信息和训练行驶环境信息；
计算单元202还用于根据模型参数和训练参数计算自动驾驶车辆的粗粒度训练 控制指令；以及，用于依据粗粒度训练控制指令计算损失函数的值；
更新单元208，用于当损失函数的值未达到预设条件时，更新模型参数；
计算单元202还用于根据更新后的模型参数和训练参数计算自动驾驶车辆的更新粗粒度训练控制指令，并重新计算损失函数的值，直到损失函数的值达到预设条件；
确定单元204还用于将损失函数的值达到预设条件时对应的模型参数确定为控制指令决策模型的最终模型参数。
可以理解的是，损失函数Loss1包括：
Loss1＝|vcosα-vsinα-Q|2，
其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预设粗粒度训练控制指令之间匹配的程度。
可以理解的是，计算单元202还用于计算模型参数的更新梯度；
更新单元208具体用于基于更新梯度、预设更新系数和更新前的模型参数，计算更新后的模型参数。
可以理解的是，判断单元203还用于判断是否执行粗粒度训练控制指令；
计算单元202具体用于：
当判定执行粗粒度训练控制指令时，通过第一预设关系计算更新梯度，第一预设关系包括更新梯度等于第一损失函数对模型参数的偏导值，第一损失函数为Loss1＝|vcosα-vsinα-Q|2，其中，v表示自动驾驶车辆的当前速度，α表示自动驾驶车辆的当前行驶方向与自动驾驶车辆所在车道之间的夹角，Q表示粗粒度训练控制指令与训练参数对应预设粗粒度训练控制指令之间匹配的程度；
当判定不执行粗粒度训练控制指令时，通过第二预设关系计算更新梯度，第二预设关系包括更新梯度等于第二损失函数对模型参数的偏导值，第二损失函数为Loss2＝|vsinα-vcosα-Q|2。
根据本发明实施例的车辆的控制装置200可对应于根据本发明实施例的车辆的控制方法中的执行主体，并且车辆的控制装置200中的各个模块的上述和其它操作和/或功能分别为了实现车辆的控制方法中的各个相应流程，为了简洁，在此不再赘述。
图6是根据本发明一实施例的车辆的控制设备300的示意性框图。如图6所示，设备300包括处理器301和存储器302总结303，总线303用于连接处理器301和存储器302，使处理器301和存储器302通过总线303进行相互通信，存储器302用于存储可执行的程序代码，处理器301通过读取存储器302中存储的可执行程序代码来运行与可执行程序代码对应的程序。
具体地，处理器301还用于执行一种车辆的控制方法；其中，车辆的控制方法包括：
在自动驾驶车辆处于自动驾驶状态时，获取自动驾驶车辆的行驶状态信息和自动驾驶车辆的第一行驶环境信息；
基于行驶状态信息、第一行驶环境信息和控制指令决策模型，计算自动驾驶车 辆的粗粒度控制指令，粗粒度控制指令用于控制自动驾驶车辆的行驶方式，控制指令决策模型为基于自动驾驶车辆在训练状态时的训练行驶状态信息和自动驾驶车辆在训练状态时的训练行驶环境信息训练得到的；
根据第二行驶环境信息判断是否执行粗粒度控制指令，第二行驶环境信息包括在自动驾驶车辆处于自动驾驶状态时，与粗粒度控制指令对应的行驶环境信息；
当判定执行粗粒度控制指令时，根据车道信息和自动驾驶车辆的行驶状态信息，确定与粗粒度控制指令对应的细粒度控制指令，细粒度控制指令用于控制自动驾驶车辆的行驶参数，车道信息包括自动驾驶车辆所行驶道路中粗粒度控制指令对应车道的信息；
输出细粒度控制指令。
根据本发明实施例的车辆的控制设备300，可对应于根据本发明实施例的车辆的控制方法中的执行主体，并且车辆的控制设备300中的各个模块的上述和其它操作和/或功能分别为了实现车辆的控制方法中的各个相应流程，为了简洁，在此不再赘述。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行所述计算机程序指令时，全部或部分地产生按照本发明实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质(例如固态硬盘(olid State Disk，SSD))等。

Claims (32)

  1. 一种车辆的控制方法，其特征在于，包括：
    在自动驾驶车辆处于自动驾驶状态时，获取所述自动驾驶车辆的行驶状态信息和所述自动驾驶车辆的第一行驶环境信息；
    基于所述行驶状态信息、所述第一行驶环境信息和控制指令决策模型，计算所述自动驾驶车辆的粗粒度控制指令，所述粗粒度控制指令用于控制所述自动驾驶车辆的行驶方式，所述控制指令决策模型为基于所述自动驾驶车辆在训练状态时的训练行驶状态信息和所述自动驾驶车辆在训练状态时的训练行驶环境信息训练得到的；
    根据第二行驶环境信息判断是否执行所述粗粒度控制指令，所述第二行驶环境信息包括在所述自动驾驶车辆处于自动驾驶状态时，与所述粗粒度控制指令对应的行驶环境信息；
    当判定执行所述粗粒度控制指令时，根据车道信息和所述自动驾驶车辆的行驶状态信息，确定与所述粗粒度控制指令对应的细粒度控制指令，所述细粒度控制指令用于控制所述自动驾驶车辆的行驶参数，所述车道信息包括所述自动驾驶车辆所行驶道路中所述粗粒度控制指令对应车道的信息；
    输出所述细粒度控制指令。
  2. 根据权利要求1所述的控制方法，其特征在于，所述粗粒度控制指令包括直行；
    所述车道信息包括所述自动驾驶车辆直行时所行驶车道的预设期望速度，所述自动驾驶车辆的行驶状态信息包括所述自动驾驶车辆的当前速度；
    所述根据车道信息和所述自动驾驶车辆的行驶状态信息，确定与所述粗粒度控制指令对应的细粒度控制指令，包括：
    基于所述预设期望速度和所述当前速度，确定所述细粒度控制指令。
  3. 根据权利要求2所述的控制方法，其特征在于，所述细粒度控制指令包括所述自动驾驶车辆的油门大小；
    所述基于所述预设期望速度和所述当前速度，确定所述细粒度控制指令，包括：
    当所述当前速度大于所述预设期望速度时，所述油门大小等于零；
    当所述当前速度不大于所述预设期望速度时，基于第一预设公式计算所述自动驾驶车辆的油门大小，其中，所述第一预设公式包括：
    所述油门大小＝预设油门控制系数×(所述预设期望速度-所述当前速度+预设值)。
  4. 根据权利要求2或3所述的控制方法，其特征在于，所述细粒度控制指令包括所述自动驾驶车辆的刹车力度；
    所述基于所述预设期望速度和所述当前速度，确定所述细粒度控制指令，包括：
    当所述当前速度小于所述预设期望速度时，所述刹车力度等于零；
    当所述当前速度不小于所述预设期望速度时，基于第二预设公式计算所述自动驾驶车辆的刹车力度，其中，所述第二预设公式包括：
    所述刹车力度＝预设刹车控制系数×(所述当前速度-所述预设期望速度)。
  5. 根据权利要求1所述的控制方法，其特征在于，所述粗粒度控制指令包括换道方向；
    所述根据车道信息和所述自动驾驶车辆的行驶状态信息，确定与所述粗粒度控制指令对应的细粒度控制指令，包括：
    基于所述自动驾驶车辆的行驶状态信息，模拟所述自动驾驶车辆在所述换道方向的换道行驶路径；
    根据所述换道行驶路径和所述车道信息，确定所述自动驾驶车辆的细粒度控制指令。
  6. 根据权利要求5所述的控制方法，其特征在于，所述细粒度控制指令包括方向盘转角；
    所述车道信息包括所述换道行驶路径对应车道的车道宽度和所述换道行驶路径对应车道的车道中线；
    所述根据所述换道行驶路径和所述车道信息，确定所述自动驾驶车辆的细粒度控制指令，包括：
    确定在所述换道行驶路径中所述自动驾驶车辆行驶方向与所述自动驾驶车辆当前所在车道直行方向之间至少一个转向角度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的车道宽度确定所述自动驾驶车辆在所述转向角度对应位置的所属车道的目标车道宽度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的车道中线确定所述自动驾驶车辆在所述转向角度对应位置与所述所属车道的车道中线之间的目标距离；
    基于第三预设公式、所述转向角度、所述目标车道宽度和所述目标距离计算所述方向盘转角，其中，所述第三预设公式包括：
    所述方向盘转角＝[所述转向角度-所述目标距离/(所述目标车道宽度×第一转角系数)]×第二转角系数。
  7. 根据权利要求6所述的控制方法，其特征在于，所述细粒度控制指令还包括所述自动驾驶车辆的油门大小；
    所述车道信息还包括所述换道行驶路径对应车道的预设期望速度；
    所述确定在所述换道行驶路径中所述自动驾驶车辆行驶方向与所述自动驾驶车辆当前所在车道直行方向之间至少一个转向角度之后，还包括：
    在所述换道行驶路径中，基于所述自动驾驶车辆的行驶状态信息确定所述自动驾驶车辆在所述转向角度对应位置的当前速度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的预设期望速度确定所述自动驾驶车辆在所述转向角度对应位置所属车道的目标预设期望速度；
    当所述当前速度大于所述目标预设期望速度时，所述油门大小等于零；
    当所述当前速度不大于所述目标预设期望速度时，基于第四预设公式计算所述自动驾驶车辆在所述转向角度对应位置的油门大小，其中，所述第四预设公式包括：
    所述油门大小＝预设油门控制系数×(所述目标预设期望速度-所述当前速度+预设值)。
  8. 根据权利要求6或7所述的控制方法，其特征在于，所述细粒度控制指令还包括所述自动驾驶车辆的刹车力度；
    所述车道信息还包括所述换道行驶路径对应车道的预设期望速度；
    所述确定在所述换道行驶路径中所述自动驾驶车辆行驶方向与所述自动驾驶车辆当前所在车道直行方向之间至少一个转向角度之后，还包括：
    在所述换道行驶路径中，基于所述自动驾驶车辆的行驶状态信息确定所述自动驾驶车辆在所述转向角度对应位置的当前速度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的预设期望速度确定所述自动驾驶车辆在所述转向角度对应位置所属车道的目标预设期望速度；
    当所述当前速度小于所述目标预设期望速度时，所述刹车力度等于零；
    当所述当前速度不小于所述目标预设期望速度时，基于第五预设公式计算所述自动驾驶车辆在所述转向角度对应位置的刹车力度，其中，所述第五预设公式包括：
    所述刹车力度＝预设刹车控制系数×(所述当前速度-所述目标预设期望速度)。
  9. 根据权利要求1-8任一项所述的控制方法，其特征在于，所述第一行驶环境信息包括所述自动驾驶车辆所行驶车道的车道信息，所述自动驾驶车辆预设距离内车辆的信息，所述自动驾驶车辆预设距离内路面的信息中至少一项；
    所述第二行驶环境信息包括所述自动驾驶车辆所行驶车道中与所述粗粒度控制指令对应的车道信息，所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的车辆信息，所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的路面信息中至少一项。
  10. 根据权利要求1-9任一项所述的控制方法，其特征在于，所述第二行驶环境信息包括目标车辆与所述自动驾驶车辆之间的第一车辆距离，所述目标车辆表示所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的车辆；
    所述根据第二行驶环境信息判断是否执行所述粗粒度控制指令，包括：
    当所述第一车辆距离大于安全距离时，判定执行所述粗粒度控制指令；
    当所述第一车辆距离不大于安全距离时，判定不执行所述粗粒度控制指令。
  11. 根据权利要求1、5-9任一项所述的控制方法，其特征在于，所述粗粒度控制指令包括换道方向；
    在所述根据第二行驶环境信息判断是否执行所述粗粒度控制指令之前，还包括：
    基于所述自动驾驶车辆的行驶状态信息模拟所述自动驾驶车辆在所述换道方向的换道行驶路径；
    所述第二行驶环境信息包括目标车辆与所述自动驾驶车辆在所述换道行驶路径行驶时的第二车辆距离，所述目标车辆表示所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的车辆；
    所述根据第二行驶环境信息判断是否执行所述粗粒度控制指令，包括：
    当所述第二车辆距离大于安全距离时，判定执行所述粗粒度控制指令；
    当所述第二车辆距离不大于安全距离时，判定不执行所述粗粒度控制指令。
  12. 根据权利要求1-11任一项所述的控制方法，其特征在于，在所述基于所述行驶状态信息、所述第一行驶环境信息和控制指令决策模型，计算所述自动驾驶车辆的粗粒度控制指令之前，还包括：
    初始化所述控制指令决策模型的模型参数；
    获取训练参数，所述训练参数包括所述训练行驶状态信息和所述训练行驶环境信 息；
    根据所述模型参数和所述训练参数计算所述自动驾驶车辆的粗粒度训练控制指令；
    依据所述粗粒度训练控制指令计算损失函数的值；
    当所述损失函数的值未达到预设条件时，更新所述模型参数；
    根据更新后的模型参数和所述训练参数计算所述自动驾驶车辆的更新粗粒度训练控制指令，并重新计算所述损失函数的值，直到所述损失函数的值达到预设条件；
    将所述损失函数的值达到预设条件时对应的模型参数确定为所述控制指令决策模型的最终模型参数。
  13. 根据权利要求12所述的控制方法，其特征在于，所述损失函数Loss1包括：
    Loss1＝|v cosα-v sinα-Q|2，
    其中，v表示所述自动驾驶车辆的当前速度，α表示所述自动驾驶车辆的当前行驶方向与所述自动驾驶车辆所在车道之间的夹角，Q表示所述粗粒度训练控制指令与所述训练参数对应预设粗粒度训练控制指令之间匹配的程度。
  14. 根据权利要求12所述的控制方法，其特征在于，所述更新所述模型参数之前，还包括：
    计算所述模型参数的更新梯度；
    所述更新所述模型参数，包括：
    基于所述更新梯度、预设更新系数和更新前的模型参数，计算更新后的模型参数。
  15. 根据权利要求14所述的控制方法，其特征在于，在所述计算所述模型参数的更新梯度之前，还包括：
    判断是否执行所述粗粒度训练控制指令；
    所述计算所述模型参数的更新梯度，包括：
    当判定执行所述粗粒度训练控制指令时，通过第一预设关系计算所述更新梯度，所述第一预设关系包括所述更新梯度等于第一损失函数对所述模型参数的偏导值，所述第一损失函数为Loss1＝|v cosα-v sinα-Q|2，其中，v表示所述自动驾驶车辆的当前速度，α表示所述自动驾驶车辆的当前行驶方向与所述自动驾驶车辆所在车道之间的夹角，Q表示所述粗粒度训练控制指令与所述训练参数对应预设粗粒度训练控制指令之间匹配的程度；
    当判定不执行所述粗粒度训练控制指令时，通过第二预设关系计算所述更新梯度，所述第二预设关系包括所述更新梯度等于第二损失函数对所述模型参数的偏导值，所述第二损失函数为Loss2＝|v sinα-v cosα-Q|2。
  16. 一种车辆的控制装置，其特征在于，包括：
    获取单元，用于在自动驾驶车辆处于自动驾驶状态时，获取所述自动驾驶车辆的行驶状态信息和所述自动驾驶车辆的第一行驶环境信息；
    计算单元，用于基于所述行驶状态信息、所述第一行驶环境信息和控制指令决策模型，计算所述自动驾驶车辆的粗粒度控制指令，所述粗粒度控制指令用于控制所述自动驾驶车辆的行驶方式，所述控制指令决策模型为基于所述自动驾驶车辆在训练状态时的训练行驶状态信息和所述自动驾驶车辆在训练状态时的训练行驶环境信息训练 得到的；
    判断单元，用于根据第二行驶环境信息判断是否执行所述粗粒度控制指令，所述第二行驶环境信息包括在所述自动驾驶车辆处于自动驾驶状态时，与所述粗粒度控制指令对应的行驶环境信息；
    确定单元，用于当判定执行所述粗粒度控制指令时，根据车道信息和所述自动驾驶车辆的行驶状态信息，确定与所述粗粒度控制指令对应的细粒度控制指令，所述细粒度控制指令用于控制所述自动驾驶车辆的行驶参数，所述车道信息包括所述自动驾驶车辆所行驶道路中所述粗粒度控制指令对应车道的信息；
    输出单元，用于输出所述细粒度控制指令。
  17. 根据权利要求16所述的控制装置，其特征在于，所述粗粒度控制指令包括直行；
    所述车道信息包括所述自动驾驶车辆直行时所行驶车道的预设期望速度，所述自动驾驶车辆的行驶状态信息包括所述自动驾驶车辆的当前速度；
    所述确定单元具体用于基于所述预设期望速度和所述当前速度，确定所述细粒度控制指令。
  18. 根据权利要求17所述的控制装置，其特征在于，所述细粒度控制指令包括所述自动驾驶车辆的油门大小；
    所述确定单元具体用于：
    当所述当前速度大于所述预设期望速度时，所述油门大小等于零；
    当所述当前速度不大于所述预设期望速度时，基于第一预设公式计算所述自动驾驶车辆的油门大小，其中，所述第一预设公式包括：
    所述油门大小＝预设油门控制系数×(所述预设期望速度-所述当前速度+预设值)。
  19. 根据权利要求17或18所述的控制装置，其特征在于，所述细粒度控制指令包括所述自动驾驶车辆的刹车力度；
    所述确定单元具体用于：
    当所述当前速度小于所述预设期望速度时，所述刹车力度等于零；
    当所述当前速度不小于所述预设期望速度时，基于第二预设公式计算所述自动驾驶车辆的刹车力度，其中，所述第二预设公式包括：
    所述刹车力度＝预设刹车控制系数×(所述当前速度-所述预设期望速度)。
  20. 根据权利要求16所述的控制装置，其特征在于，所述粗粒度控制指令包括换道方向；
    所述确定单元具体用于：
    基于所述自动驾驶车辆的行驶状态信息，模拟所述自动驾驶车辆在所述换道方向的换道行驶路径；
    根据所述换道行驶路径和所述车道信息，确定所述自动驾驶车辆的细粒度控制指令。
  21. 根据权利要求20所述的控制装置，其特征在于，所述细粒度控制指令包括方向盘转角；
    所述车道信息包括所述换道行驶路径对应车道的车道宽度和所述换道行驶路径对 应车道的车道中线；
    所述确定单元具体用于：
    确定在所述换道行驶路径中所述自动驾驶车辆行驶方向与所述自动驾驶车辆当前所在车道直行方向之间至少一个转向角度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的车道宽度确定所述自动驾驶车辆在所述转向角度对应位置的所属车道的目标车道宽度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的车道中线确定所述自动驾驶车辆在所述转向角度对应位置与所述所属车道的车道中线之间的目标距离；
    基于第三预设公式、所述转向角度、所述目标车道宽度和所述目标距离计算所述方向盘转角，其中，所述第三预设公式包括：
    所述方向盘转角＝[所述转向角度-所述目标距离/(所述目标车道宽度×第一转角系数)]×第二转角系数。
  22. 根据权利要求21所述的控制装置，其特征在于，所述细粒度控制指令还包括所述自动驾驶车辆的油门大小；
    所述车道信息还包括所述换道行驶路径对应车道的预设期望速度；
    所述确定单元具体用于：
    在所述换道行驶路径中，基于所述自动驾驶车辆的行驶状态信息确定所述自动驾驶车辆在所述转向角度对应位置的当前速度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的预设期望速度确定所述自动驾驶车辆在所述转向角度对应位置所属车道的目标预设期望速度；
    当所述当前速度大于所述目标预设期望速度时，所述油门大小等于零；
    当所述当前速度不大于所述目标预设期望速度时，基于第四预设公式计算所述自动驾驶车辆在所述转向角度对应位置的油门大小，其中，所述第四预设公式包括：
    所述油门大小＝预设油门控制系数×(所述目标预设期望速度-所述当前速度+预设值)。
  23. 根据权利要求21或22所述的控制装置，其特征在于，所述细粒度控制指令还包括所述自动驾驶车辆的刹车力度；
    所述车道信息还包括所述换道行驶路径对应车道的预设期望速度；
    所述确定单元具体用于：
    在所述换道行驶路径中，基于所述自动驾驶车辆的行驶状态信息确定所述自动驾驶车辆在所述转向角度对应位置的当前速度；
    在所述换道行驶路径中，根据所述换道行驶路径对应车道的预设期望速度确定所述自动驾驶车辆在所述转向角度对应位置所属车道的目标预设期望速度；
    当所述当前速度小于所述目标预设期望速度时，所述刹车力度等于零；
    当所述当前速度不小于所述目标预设期望速度时，基于第五预设公式计算所述自动驾驶车辆在所述转向角度对应位置的刹车力度，其中，所述第五预设公式包括：
    所述刹车力度＝预设刹车控制系数×(所述当前速度-所述目标预设期望速度)。
  24. 根据权利要求16-23任一项所述的控制装置，其特征在于，所述第一行驶环境信息包括所述自动驾驶车辆所行驶车道的车道信息，所述自动驾驶车辆预设距离内车 辆的信息，所述自动驾驶车辆预设距离内路面的信息中至少一项；
    所述第二行驶环境信息包括所述自动驾驶车辆所行驶车道中与所述粗粒度控制指令对应的车道信息，所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的车辆信息，所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的路面信息中至少一项。
  25. 根据权利要求16-24任一项所述的控制装置，其特征在于，所述第二行驶环境信息包括目标车辆与所述自动驾驶车辆之间的第一车辆距离，所述目标车辆表示所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的车辆；
    所述判断单元具体用于：
    当所述第一车辆距离大于安全距离时，判定执行所述粗粒度控制指令；
    当所述第一车辆距离不大于安全距离时，判定不执行所述粗粒度控制指令。
  26. 根据权利要求16、20-24任一项所述的控制装置，其特征在于，所述粗粒度控制指令包括换道方向；所述装置还包括：
    模拟单元，用于基于所述自动驾驶车辆的行驶状态信息模拟所述自动驾驶车辆在所述换道方向的换道行驶路径；
    所述第二行驶环境信息包括目标车辆与所述自动驾驶车辆在所述换道行驶路径行驶时的第二车辆距离，所述目标车辆表示所述自动驾驶车辆预设距离内与所述粗粒度控制指令对应的车辆；
    所述判断单元具体用于：
    当所述第二车辆距离大于安全距离时，判定执行所述粗粒度控制指令；
    当所述第二车辆距离不大于安全距离时，判定不执行所述粗粒度控制指令。
  27. 根据权利要求16-26任一项所述的控制装置，其特征在于，还包括：
    初始化单元，用于初始化所述控制指令决策模型的模型参数；
    所述获取单元还用于获取训练参数，所述训练参数包括所述训练行驶状态信息和所述训练行驶环境信息；
    所述计算单元还用于根据所述模型参数和所述训练参数计算所述自动驾驶车辆的粗粒度训练控制指令；以及，用于依据所述粗粒度训练控制指令计算损失函数的值；
    所述装置还包括：
    更新单元，用于当所述损失函数的值未达到预设条件时，更新所述模型参数；
    所述计算单元，还用于根据更新后的模型参数和所述训练参数计算所述自动驾驶车辆的更新粗粒度训练控制指令，并重新计算所述损失函数的值，直到所述损失函数的值达到预设条件；
    所述确定单元还用于将所述损失函数的值达到预设条件时对应的模型参数确定为所述控制指令决策模型的最终模型参数。
  28. 根据权利要求27所述的控制装置，其特征在于，所述损失函数Loss1包括：
    Loss1＝|v cosα-v sinα-Q|2，
    其中，v表示所述自动驾驶车辆的当前速度，α表示所述自动驾驶车辆的当前行驶方向与所述自动驾驶车辆所在车道之间的夹角，Q表示所述粗粒度训练控制指令与所述训练参数对应预设粗粒度训练控制指令之间匹配的程度。
  29. 根据权利要求27所述的控制装置，其特征在于，所述计算单元还用于计算所述模型参数的更新梯度；
    所述更新单元具体用于基于所述更新梯度、预设更新系数和更新前的模型参数，计算更新后的模型参数。
  30. 根据权利要求29所述的控制装置，其特征在于，所述判断单元还用于判断是否执行所述粗粒度训练控制指令；
    所述计算单元具体用于：
    当判定执行所述粗粒度训练控制指令时，通过第一预设关系计算所述更新梯度，所述第一预设关系包括所述更新梯度等于第一损失函数对所述模型参数的偏导值，所述第一损失函数为Loss1＝|v cosα-v sinα-Q|2，其中，v表示所述自动驾驶车辆的当前速度，α表示所述自动驾驶车辆的当前行驶方向与所述自动驾驶车辆所在车道之间的夹角，Q表示所述粗粒度训练控制指令与所述训练参数对应预设粗粒度训练控制指令之间匹配的程度；
    当判定不执行所述粗粒度训练控制指令时，通过第二预设关系计算所述更新梯度，所述第二预设关系包括所述更新梯度等于第二损失函数对所述模型参数的偏导值，所述第二损失函数为Loss2＝|v sinα-v cosα-Q|2。
  31. 一种车辆的控制设备，其特征在于，包括：
    存储器、处理器和总线；
    所述存储器和所述处理器通过所述总线连接并完成相互间的通信；
    所述存储器用于存储程序代码；
    所述处理器通过读取所述存储器中存储的可执行程序代码来运行与所述可执行程序代码对应的程序，以用于执行如权利要求1-15任一项所述的控制方法。
  32. 一种计算机可读存储介质，其特征在于，包括指令，当所述指令在计算机上运行时，使得所述计算机执行如权利要求1-15任一项所述的方法。
PCT/CN2017/091095 2017-06-30 2017-06-30 车辆的控制方法、装置及设备 WO2019000391A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
PCT/CN2017/091095 WO2019000391A1 (zh) 2017-06-30 2017-06-30 车辆的控制方法、装置及设备
EP17915345.7A EP3647140B1 (en) 2017-06-30 2017-06-30 Vehicle control method, device, and apparatus
CN201780092791.7A CN110809542B (zh) 2017-06-30 2017-06-30 车辆的控制方法、装置及设备
US16/730,444 US20200139989A1 (en) 2017-06-30 2019-12-30 Vehicle Control Method, Apparatus, and Device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/091095 WO2019000391A1 (zh) 2017-06-30 2017-06-30 车辆的控制方法、装置及设备

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/730,444 Continuation US20200139989A1 (en) 2017-06-30 2019-12-30 Vehicle Control Method, Apparatus, and Device

Publications (1)

Publication Number Publication Date
WO2019000391A1 true WO2019000391A1 (zh) 2019-01-03

Family

ID=64740754

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/091095 WO2019000391A1 (zh) 2017-06-30 2017-06-30 车辆的控制方法、装置及设备

Country Status (4)

Country Link
US (1) US20200139989A1 (zh)
EP (1) EP3647140B1 (zh)
CN (1) CN110809542B (zh)
WO (1) WO2019000391A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200102378A (ko) * 2019-02-21 2020-08-31 바이두 온라인 네트웍 테크놀러지 (베이징) 캄파니 리미티드 정보 처리 방법, 장치 및 저장 매체
WO2021051959A1 (zh) * 2019-09-16 2021-03-25 华为技术有限公司 车辆控制的方法、装置、控制器和智能汽车

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11130497B2 (en) 2017-12-18 2021-09-28 Plusai Limited Method and system for ensemble vehicle control prediction in autonomous driving vehicles
US11273836B2 (en) * 2017-12-18 2022-03-15 Plusai, Inc. Method and system for human-like driving lane planning in autonomous driving vehicles
US20190185012A1 (en) 2017-12-18 2019-06-20 PlusAI Corp Method and system for personalized motion planning in autonomous driving vehicles
US10940863B2 (en) * 2018-11-01 2021-03-09 GM Global Technology Operations LLC Spatial and temporal attention-based deep reinforcement learning of hierarchical lane-change policies for controlling an autonomous vehicle
US11079761B2 (en) * 2018-12-12 2021-08-03 Ford Global Technologies, Llc Vehicle path processing
US20200406894A1 (en) * 2019-06-28 2020-12-31 Zoox, Inc. System and method for determining a target vehicle speed
CN112052183B (zh) * 2020-09-28 2023-07-07 英博超算(南京)科技有限公司 一种基于虚拟车辆平台的中间件调试方法
CN112904852B (zh) * 2021-01-18 2021-09-17 国汽智控(北京)科技有限公司 一种自动驾驶控制方法、装置及电子设备
CN113264064B (zh) * 2021-03-31 2022-05-10 志行千里(北京)科技有限公司 用于交叉路口场景的自动驾驶方法及相关设备
KR102331807B1 (ko) * 2021-04-16 2021-12-02 국방과학연구소 차량의 주행을 위한 전자 장치 및 그의 동작 방법
KR20230146076A (ko) * 2021-09-10 2023-10-18 후아웨이 테크놀러지 컴퍼니 리미티드 지오펜스 주행 정책을 획득하기 위한 시뮬레이션 기반 방법 및 데이터 센터
KR102573526B1 (ko) * 2022-07-08 2023-09-06 주식회사 노타 강화학습 모델을 이용하여 연동 그룹 내 신호기의 교통신호를 제어하는 장치 및 방법

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104635736A (zh) * 2014-12-19 2015-05-20 财团法人车辆研究测试中心 用于驾驶行为决策的自动驾驶系统及其方法
WO2015145606A1 (ja) * 2014-03-26 2015-10-01 日産自動車株式会社 情報呈示装置及び情報呈示方法
CN205451512U (zh) * 2016-01-05 2016-08-10 常州加美科技有限公司 一种交通信号灯路口无人驾驶汽车安全行驶装置
CN105892471A (zh) * 2016-07-01 2016-08-24 北京智行者科技有限公司 汽车自动驾驶方法和装置
CN106103231A (zh) * 2014-03-18 2016-11-09 日产自动车株式会社 车辆操作装置

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4597465A (en) * 1982-10-27 1986-07-01 A.R.A. Manufacturing Company Cruise control system and method with overspeed sensor
US6393362B1 (en) * 2000-03-07 2002-05-21 Modular Mining Systems, Inc. Dynamic safety envelope for autonomous-vehicle collision avoidance system
JP4525670B2 (ja) * 2006-11-20 2010-08-18 トヨタ自動車株式会社 走行制御計画生成システム
US20100228427A1 (en) * 2009-03-05 2010-09-09 Massachusetts Institute Of Technology Predictive semi-autonomous vehicle navigation system
KR101553190B1 (ko) * 2009-08-18 2015-09-14 도요타 지도샤（주） 차량 제어 시스템
JP5625515B2 (ja) * 2010-06-10 2014-11-19 株式会社デンソー 車両制駆動制御装置
GB201305067D0 (en) * 2013-03-19 2013-05-01 Massive Analytic Ltd Apparatus for controlling a land vehicle which is self-driving or partially self-driving
JP6025273B2 (ja) * 2015-03-17 2016-11-16 富士重工業株式会社 車両の走行制御装置
US9934688B2 (en) * 2015-07-31 2018-04-03 Ford Global Technologies, Llc Vehicle trajectory determination
US10249194B2 (en) * 2016-08-30 2019-04-02 International Business Machines Corporation Modifying behavior of autonomous vehicle based on advanced predicted behavior analysis of nearby drivers
US10606276B2 (en) * 2016-09-30 2020-03-31 Faraday & Future Inc. User data-based autonomous vehicle system
US10229600B2 (en) * 2017-02-22 2019-03-12 Telenav, Inc. Navigation system with traffic flow mechanism and method of operation thereof
US10870426B2 (en) * 2017-06-22 2020-12-22 Magna Electronics Inc. Driving assistance system with rear collision mitigation

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106103231A (zh) * 2014-03-18 2016-11-09 日产自动车株式会社 车辆操作装置
WO2015145606A1 (ja) * 2014-03-26 2015-10-01 日産自動車株式会社 情報呈示装置及び情報呈示方法
CN104635736A (zh) * 2014-12-19 2015-05-20 财团法人车辆研究测试中心 用于驾驶行为决策的自动驾驶系统及其方法
CN205451512U (zh) * 2016-01-05 2016-08-10 常州加美科技有限公司 一种交通信号灯路口无人驾驶汽车安全行驶装置
CN105892471A (zh) * 2016-07-01 2016-08-24 北京智行者科技有限公司 汽车自动驾驶方法和装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3647140A4 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200102378A (ko) * 2019-02-21 2020-08-31 바이두 온라인 네트웍 테크놀러지 (베이징) 캄파니 리미티드 정보 처리 방법, 장치 및 저장 매체
EP3699890A3 (en) * 2019-02-21 2020-11-25 Baidu Online Network Technology (Beijing) Co., Ltd. Information processing method and apparatus, and storage medium
KR102424067B1 (ko) * 2019-02-21 2022-07-22 아폴로 인텔리전트 드라이빙 테크놀로지(베이징) 컴퍼니 리미티드 정보 처리 방법, 장치 및 저장 매체
WO2021051959A1 (zh) * 2019-09-16 2021-03-25 华为技术有限公司 车辆控制的方法、装置、控制器和智能汽车

Also Published As

Publication number Publication date
EP3647140B1 (en) 2022-07-20
CN110809542A (zh) 2020-02-18
CN110809542B (zh) 2021-05-11
EP3647140A4 (en) 2020-07-29
EP3647140A1 (en) 2020-05-06
US20200139989A1 (en) 2020-05-07

Similar Documents

Publication Publication Date Title
WO2019000391A1 (zh) 车辆的控制方法、装置及设备
CN112733270B (zh) 车辆行驶轨迹预测和轨迹偏离危险度评估的系统与方法
US11493926B2 (en) Offline agent using reinforcement learning to speedup trajectory planning for autonomous vehicles
WO2020011025A1 (zh) 一种车辆的自动换道方法及装置
US10800427B2 (en) Systems and methods for a vehicle controller robust to time delays
WO2018176593A1 (zh) 一种面向无人自行车的局部避障路径规划方法
JP2020001678A (ja) 重み付け幾何学的コストを有する区分的螺旋曲線を使用した基準線平滑化方法
CN111679660B (zh) 一种融合类人驾驶行为的无人驾驶深度强化学习方法
CN111696339B (zh) 一种自动驾驶车队的跟车控制方法、系统及车辆
CN112327830B (zh) 车辆自动驾驶变道轨迹的规划方法及电子设备
CN114265411B (zh) 一种解决车辆预测模型性能受限于感知数据性能的方法
Liu et al. Real-time longitudinal and lateral state estimation of preceding vehicle based on moving horizon estimation
US20220155732A9 (en) System and Method of Efficient, Continuous, and Safe Learning Using First Principles and Constraints
CN114771551B (zh) 自动驾驶车辆轨迹规划方法、装置和自动驾驶车辆
US11780464B2 (en) Autonomous vehicle trajectory generation using velocity-based steering limits
CN110525421B (zh) 一种用于具有挂车的车辆的车道保持强化学习方法及系统
CN115617051A (zh) 车辆控制方法、装置、设备和计算机可读介质
Xiong et al. Surrounding vehicle trajectory prediction and dynamic speed planning for autonomous vehicle in cut-in scenarios
Nan et al. Interaction-aware planning with deep inverse reinforcement learning for human-like autonomous driving in merge scenarios
CN109492835A (zh) 车辆控制信息的确定方法、模型训练方法和相关装置
CN115476881B (zh) 车辆轨迹跟踪控制方法、装置、设备及介质
CN114906168B (zh) 一种车辆自动驾驶的控制方法、系统、设备及计算机可读存储介质
CN115447615A (zh) 基于车辆运动学模型预测控制的轨迹优化方法
CN115542899A (zh) 车辆路径跟踪的方法、装置、车辆、电子设备及介质
Wang et al. An end-to-end deep reinforcement learning model based on proximal policy optimization algorithm for autonomous driving of off-road vehicle

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17915345

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2017915345

Country of ref document: EP

Effective date: 20200129