WO2021017504A1 - 运动轨迹预测方法和装置 - Google Patents

运动轨迹预测方法和装置 Download PDF

Info

Publication number
WO2021017504A1
WO2021017504A1 PCT/CN2020/081894 CN2020081894W WO2021017504A1 WO 2021017504 A1 WO2021017504 A1 WO 2021017504A1 CN 2020081894 W CN2020081894 W CN 2020081894W WO 2021017504 A1 WO2021017504 A1 WO 2021017504A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
motion
state
initial state
states
Prior art date
Application number
PCT/CN2020/081894
Other languages
English (en)
French (fr)
Inventor
曹彤彤
刘亚林
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 EP20846548.4A priority Critical patent/EP3992732A4/en
Publication of WO2021017504A1 publication Critical patent/WO2021017504A1/zh
Priority to US17/585,802 priority patent/US20220144265A1/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
    • 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
    • B60W30/10—Path keeping
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
    • 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
    • B60W50/0097—Predicting future conditions
    • 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
    • B60W60/0027—Planning or execution of driving tasks using trajectory prediction for other traffic participants
    • B60W60/00274—Planning or execution of driving tasks using trajectory prediction for other traffic participants considering possible movement changes
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B17/00—Systems involving the use of models or simulators of said systems
    • G05B17/02—Systems involving the use of models or simulators of said systems electric
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • 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
    • 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
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • 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
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
    • 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
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
    • G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
    • G05D1/0248—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means in combination with a laser
    • 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
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0255—Control of position or course in two dimensions specially adapted to land vehicles using acoustic signals, e.g. ultra-sonic singals
    • 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
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0257—Control of position or course in two dimensions specially adapted to land vehicles using a radar
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
    • G08G1/0112—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from the vehicle, e.g. floating car data [FCD]
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/04—Detecting movement of traffic to be counted or controlled using optical or ultrasonic detectors
    • 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
    • B60W2520/00—Input parameters relating to overall vehicle dynamics
    • B60W2520/10—Longitudinal speed
    • B60W2520/105—Longitudinal acceleration
    • 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
    • B60W2552/30—Road curve radius
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/66—Radar-tracking systems; Analogous systems
    • G01S13/72—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar
    • G01S13/723—Radar-tracking systems; Analogous systems for two-dimensional tracking, e.g. combination of angle and range tracking, track-while-scan radar by using numerical data
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
    • G01S13/931—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of land vehicles
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
    • G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Definitions

  • the invention relates to a motion trajectory prediction technology, in particular to a motion trajectory prediction method and device.
  • Intelligent driving systems usually include modules such as perception, positioning, and regulation.
  • the sensing module is used to perceive the environment around the vehicle
  • the positioning module is used to determine the location of the vehicle
  • the regulation module is used to combine the sensing result of the sensing module and the positioning result of the positioning module to plan and control the intelligent driving vehicle.
  • an intelligent driving vehicle not only needs to fully perceive the surrounding environment of the vehicle itself, but also needs to predict its future trajectory changes in advance to avoid possible collisions.
  • Target trajectory prediction is often located in the control module. It predicts the future trajectory of the target based on the perception of the target. The correct trajectory prediction can provide important reference information for the path planning of the control module.
  • the prior art usually uses a certain kinematic model (such as constant velocity, constant acceleration or constant angular velocity) to predict the trajectory based on the position and velocity of the target.
  • a certain kinematic model such as constant velocity, constant acceleration or constant angular velocity
  • the solid line represents the historical trajectory of the target
  • the dashed line represents the predicted trajectory inferred from the target's current position and speed. Since the target's future movement will not completely follow the original movement characteristics, especially when its future movement is adjusted according to the road structure, there is a large deviation in the trajectory prediction using the existing technology.
  • embodiments of the present invention provide a motion trajectory prediction method and device for improving the accuracy of motion trajectory prediction.
  • the embodiments of the present invention can be applied to an intelligent driving system including automatic driving and assisted driving.
  • the target may not only be a vehicle target, but also a pedestrian, a cyclist, a robot, or other moving targets.
  • a motion trajectory prediction method including: acquiring an initial state of a target, the initial state including an initial position and an initial motion state; and generating one or more of the targets according to the initial state and preset path information A plurality of target states, the target state includes at least a target position; a motion trajectory prediction is performed on the target according to the initial state and the target state to obtain one or more predicted motion trajectories.
  • preset path information in the form of road structures, road signs, or maps
  • the predicted trajectory of the target is constrained by the preset path information. Compared with the trajectory predicted by the initial state of the target alone, it is more in line with the moving target, especially The actual motion scene of the vehicle target can greatly improve the accuracy of the motion trajectory prediction, and it can be predicted for a long time (such as 10s) trajectory.
  • the generating one or more destination states of the target according to the initial state and preset path information includes: The initial position and the preset path information determine one or more preset paths associated with the target; the one or more destination states are generated according to the initial state and the motion model, the one or more Each destination state corresponds to the one or more preset paths. Determining the preset path associated with the target helps to associate the initial state of the target with the preset path information, so as to obtain the destination state associated with the preset path information.
  • the determined preset path associated with the target may be a preset path closest to the target, or may be a plurality of preset paths adjacent to the target.
  • the one or more are generated according to the initial state and the motion model
  • the target states include: establishing one or more Frenet coordinate systems based on the one or more preset paths; respectively, in the one or more Frenet coordinate systems, generating the target according to the initial state and motion model of the target One or more destination states.
  • the method further includes: converting the one or more predicted motion trajectories into a coordinate system used for target motion planning control.
  • the predicted motion trajectory is converted from the Frenet coordinate system to the coordinate system used for target motion planning control, which is convenient for applying the target motion trajectory prediction result to the motion planning control module.
  • the coordinate system used for the target motion planning control includes a vehicle body coordinate system , Station center coordinate system or world coordinate system.
  • the generating of the motion path is based on the initial state and preset path information.
  • the one or more goal states of the target include: generating one or more goal states of the target according to the initial state, preset path information, and parameters after model training.
  • the target state generated by the parameters after model training has a higher matching degree with the actual target motion characteristics, and the prediction of the trajectory is more accurate.
  • the method further includes: generating a probability of each of the multiple destination states according to the initial state of the target and the preset path information, and the probability of each destination state is used to obtain multiple predictions The probability of each of the trajectories.
  • the probability of each of the trajectories In the actual motion environment, there may be multiple possible trajectories of the target in the next step. By introducing probability calculation, it is helpful to make better judgments on the possible choices of the target, and further improve the accuracy of motion trajectory prediction.
  • Predicting the motion trajectory of the target to obtain one or more predicted motion trajectories includes: predicting the motion trajectory of the target according to the initial state and the target state with the greatest probability to obtain the predicted motion trajectory with the greatest probability.
  • the target state further includes a target motion state
  • the target motion state includes At least one of velocity, acceleration, angular velocity, or angular acceleration. Combining the target position and target motion state to predict the motion trajectory can further improve the accuracy of the prediction.
  • the initial motion state includes at least one of velocity, acceleration, angular velocity, or angular acceleration One.
  • the initial state of the target and the preset path information are based on the same
  • the obtaining the initial state of the target includes: obtaining the target perception result from one or more sensors; and projecting the target perception result into the same coordinate system to obtain the initial state of the target.
  • the result output by the sensor is usually based on a different coordinate system from the preset path information, and the two are mapped to the same coordinate system through a projection operation, which is convenient for the subsequent operations to integrate the two to generate the target state of the target.
  • the target perception result is based on a sensor coordinate system or a vehicle body coordinate system,
  • the same coordinate system is the station center coordinate system or the world coordinate system.
  • a motion trajectory prediction device including: an initial state acquisition module for acquiring an initial state of a target, the initial state including an initial position and an initial motion state; and a target state generation module for acquiring an initial state according to the initial state State and preset path information to generate one or more target states of the target, the target state includes at least a target position; a motion trajectory prediction module is used to perform the target state according to the initial state and the target state Motion trajectory prediction, to obtain one or more predicted motion trajectories.
  • the target state generation module includes: an associated path determination module, configured to determine the path according to the initial position and the preset path information Determine one or more preset paths associated with the target; a model calculation module for generating the one or more destination states according to the initial state and the motion model, the one or more destination states corresponding to The one or more preset paths.
  • the model calculation module includes: a coordinate system establishment module for The one or more preset paths establish one or more Frenet coordinate systems; the one or more target states are respectively determined by the model calculation module according to the initial state of the target in the one or more Frenet coordinate systems And motion model generation.
  • the one or more predicted motion trajectories are in the one or A plurality of motion trajectories in Frenet coordinate systems, and the device further includes a coordinate system conversion module for converting the one or more predicted motion trajectories into a coordinate system for target motion planning control.
  • the coordinate system used for the target motion planning control includes a vehicle body coordinate system , Station center coordinate system or world coordinate system.
  • the generating of the motion path is based on the initial state and preset path information.
  • the one or more goal states of the target include: generating one or more goal states of the target according to the initial state, preset path information, and parameters after model training.
  • the target state generation module includes a probability calculation module for when the When one or more destination states are multiple destination states, the probability of generating each of the multiple destination states according to the initial state of the target and the preset path information, the The probability is used to obtain the probability of each of the multiple predicted motion trajectories.
  • the initial state and the one or more target states Predicting the motion trajectory of the target to obtain one or more predicted motion trajectories includes: predicting the motion trajectory of the target according to the initial state and the target state with the greatest probability to obtain the predicted motion trajectory with the greatest probability.
  • the target state further includes a target motion state
  • the target motion state includes At least one of velocity, acceleration, angular velocity, or angular acceleration.
  • the initial motion state includes at least one of velocity, acceleration, angular velocity, or angular acceleration One.
  • the acquiring the initial state of the target includes: acquiring the target perception result from one or more sensors; and projecting the target perception result into the same coordinate system to obtain the initial state of the target .
  • the result output by the sensor is usually based on a different coordinate system from the preset path information, and the two are mapped to the same coordinate system through a projection operation, which is convenient for the subsequent operations to integrate the two to generate the target state of the target.
  • the target perception result is based on a sensor coordinate system or a vehicle body coordinate system,
  • the same coordinate system is the station center coordinate system or the world coordinate system.
  • a motion trajectory prediction device including a memory and a processor, the memory stores computer program instructions, and the processor runs the computer program instructions to execute the method according to any one of claims 1-8 .
  • a computer storage medium including computer instructions, which when the computer instructions are executed by a processor, cause the motion trajectory prediction apparatus to execute the method according to any one of claims 1-8.
  • a computer program product which when the computer program product runs on a processor, causes the motion trajectory prediction device to execute the method according to any one of claims 1-8.
  • Figure 1 is a schematic diagram of a prior art motion trajectory prediction method
  • FIG. 2 is a schematic diagram of the Frenet coordinate system in an embodiment of the present invention.
  • Figure 3 is a diagram of a typical application scenario of an embodiment of the present invention (taking a driving scenario of a motor vehicle on a road as an example);
  • Embodiment 4 is a flowchart of a motion trajectory prediction method provided by Embodiment 1 of the present invention.
  • Embodiment 1 of the present invention is a schematic diagram of a situation where one target corresponds to multiple preset paths in Embodiment 1 of the present invention
  • FIG. 6 is a schematic diagram of transforming the movement track of the target from the Frenet coordinate system to the station center coordinate system in the first embodiment of the present invention
  • FIG. 7 is a structural block diagram of a motion trajectory prediction device provided by Embodiment 2 of the present invention.
  • FIG. 8 is a structural block diagram of a motion trajectory prediction device provided in Embodiment 3 of the present invention.
  • Sensor coordinate system The coordinate system on which the sensing result of each sensor is based. Depending on the sensor type or installation location, the coordinate system used by each sensor may also be different.
  • Vehicle body coordinate system used to describe the relative positional relationship between objects around the vehicle and the vehicle.
  • the three mutually perpendicular coordinate axes are the length, width, and height of the vehicle.
  • SAE Society of Automotive Engineers
  • Station center coordinate system also known as ENU coordinate system or NEU coordinate system or northeast sky coordinate system.
  • the three coordinate axes point to mutually perpendicular east, north and sky respectively.
  • the world coordinate system is a coordinate system that describes the positional relationship on the earth.
  • Common world coordinate systems include WGS-84 coordinate system, UTM (Universal Transverse Mercator, Universal Transverse Mercator) coordinate system.
  • Frenet coordinate system As shown in Figure 2, the center line of the preset path is used as the reference line, the origin of the coordinate system is defined as a reference point on the reference line, and the s axis of the coordinate system is the moving target along the The forward direction of the reference line, the d axis of the coordinate system is the connection direction between the projection point of the moving target on the reference line and the target, and the connection between the moving target and the projection point is perpendicular to The tangent of the reference line at the projection point.
  • Fig. 3 uses the driving scene of a motor vehicle on the road as an example to introduce a typical application scenario diagram of the embodiment of the present invention, but the application scenario of the embodiment of the present invention is not limited to this specific situation, and can also be applied to all moving targets (including vehicles).
  • the subsequent embodiments take vehicles or lanes as examples for description, but this field Technicians can extend it to the field of motion trajectory prediction for other targets.
  • the target vehicle travels along the right side of the lane centerline at position A at speed V 0 (the speed in the embodiment of the present invention includes the speed and direction). Divide into two (respectively go straight and turn right), so the target vehicle is likely to appear at position B or position C at time t to be predicted, so there are two trajectories of the target vehicle from time t 0 to time t. It is possible to go side by side, that is, continue straight (driving from position A to position B) or turn right (driving from position A to position C).
  • Fig. 4 is a flowchart of an embodiment of the motion trajectory prediction method of the present invention. As shown in Figure 4, the motion trajectory prediction method includes step 401, step 402, and step 403:
  • Step 401 Obtain the initial state of the target.
  • the targets include pedestrians, motor vehicles or non-motor vehicles, and may also include other moving targets.
  • step 401 includes a coordinate conversion operation to project the target perception result into the coordinate based on the preset path information, and obtain the target based on the same coordinate system as the preset path information.
  • the initial state of the target includes an initial position and an initial motion state, and the initial motion state includes velocity, acceleration, angular velocity, or angular acceleration.
  • Step 402 Generate one or more destination states of the target according to the initial state and preset path information.
  • the preset path information may be information in the form of road structure, road sign indication or map.
  • the target state includes at least a target position, and may also include a target motion state, such as velocity, acceleration, angular velocity, or angular acceleration. Combining the target position and target motion state to predict the motion trajectory can further improve the accuracy of the prediction.
  • Step 402 includes: determining one or more preset paths associated with the target according to the initial position information in the initial state of the target and the preset path information. Specifically, in the case shown in FIG. 3 as an example, can find the path information in the form of a road traffic, etc predetermined time t 0 in the initial position of the target vehicle, obtaining the target vehicle at time t 0 corresponding to the center of the lane and the lane The lane and the center line of the lane are the preset path related to the target vehicle in Figure 3.
  • FIG. 5 is a schematic diagram of the situation where one target corresponds to multiple preset paths.
  • the driving scenario above corresponds to the application scenario for example.
  • the starting point A is the initial position of the target vehicle
  • the speed V 0 is the initial motion state of the target vehicle, and both are included in the initial state of the target vehicle. Since the target vehicle is located in the lane where the first lane and the second lane overlap at the initial time t 0 , the centerline of the first lane and the centerline of the second lane can be used as multiple preset paths associated with the target vehicle.
  • the preset path closest to the initial position of the moving target can be selected as the associated preset path, or multiple preset paths whose distance to the moving target is within a set threshold can be selected as the associated preset path.
  • the one or more preset paths that best match the direction of acceleration are used as the associated preset paths.
  • the probability of the destination state corresponding to each associated preset path can be obtained by calculating the probability of each associated preset path, so The probability of the stated state is used to calculate the probability of the predicted motion trajectory, which can provide a reference for subsequent more accurate regulatory decisions.
  • Step 402 also includes: establishing one or more Frenet coordinate systems based on the selected one or more preset paths, respectively, in the one or more Frenet coordinate systems, generating the target according to the initial state and motion model of the target One or more destination states.
  • two destination states are generated according to the initial state and motion model of the target vehicle.
  • the two destination states respectively correspond to the two previously selected associated preset paths (ie, the centerline of the first lane).
  • the destination state corresponding to the center line of the first lane is the destination point B and the speed V t
  • the destination state corresponding to the center line of the second lane is the destination C and the speed V t ′.
  • the s-axis of the Frenet coordinate system is the target along the associated preset path (for example, the centerline of the first lane or the centerline of the second lane in Figure 5).
  • the d-axis of the coordinate system is the connection direction between the projection point of the target on the associated preset path and the target, and the connection between the target and the projection point is perpendicular to the target
  • the tangent of the associated preset path at the projection point, with the projection point as the origin, the initial position of the target at time 0 is P 0 (the components of P 0 on the s axis and the d axis are P 0S and P 0d )
  • the initial velocity of the target is V 0 (the components of V 0 on the s axis and the d axis are V 0S and V 0d respectively )
  • the target position at time T to be predicted is P T (the components of P T on the s axis and d axis P TS and P Td respectively )
  • the target velocity of the target is V T (the components of V T on the s-axis and d-axis are V TS and V Td ).
  • the motion models used on the s-axis and d-axis are respectively a uniform motion model and a uniform deceleration motion model, and assuming that the speed direction of the target at time T is parallel to the associated preset path, then the time at T to be predicted
  • the destination status is:
  • V TS V 0S ;
  • V Td 0.
  • the target state at time T to be predicted is:
  • V TS V 0S +aT
  • V Td V 0d .
  • a target may be associated with multiple preset paths, corresponding to multiple Frenet coordinate systems, and multiple corresponding destination states will also be generated.
  • the probability of each destination state is also generated according to the initial state of the target.
  • the probabilities of the destination location B and the destination location C can be generated separately, and the two probabilities together with the two destination states can be provided to the path planning and control module to better assist the regulatory decision-making; or
  • the destination state with the highest probability can be determined by the above two probabilities, and only the destination state with the highest probability is provided to the path planning and control module.
  • step 402 "generate one or more destination states of the target according to the initial state and preset path information" also adopts the parameters after model training.
  • Model training refers to the use of existing sample data to determine the parameters of the model through some methods. The following three examples are used to illustrate how this embodiment uses the parameters after model training:
  • the distance between the initial position of the target and each lane and the angle between the initial speed direction and the lane line can be combined to establish a multi-classifier model, and the corresponding lanes can be obtained through algorithms such as softmax, boosting or decision trees.
  • the probability Taking softmax as an example, the following model is established based on the distance between the initial target position and each lane and the angle between the initial speed direction and the lane line:
  • w is the weight used to calculate the probability that the target is associated with the corresponding lane
  • d 0 is the distance between the target's initial position and the lane (that is, the distance between the projection of the target's initial position to the centerline of the lane and the target's initial position)
  • ⁇ h 0 is the angle between the initial speed direction of the target and the direction of the lane centerline
  • alpha1 and alpha2 are the parameters after model training
  • p j is the probability that the target is associated with lane j
  • j is an integer
  • w i is the probability that the target is associated with lane i
  • the weight of i needs to traverse all lanes.
  • l is the length of the predicted trajectory, which is determined by the s component of the destination position in the Frenet coordinate system
  • d 0 is the d component of the initial position in the Frenet coordinate system
  • alpha and beta are the parameters after model training, which can be trained by linear regression .
  • the following displacement attenuation model can be constructed
  • l is the length of the predicted trajectory, which is determined by the s component of the destination position in the Frenet coordinate system
  • v 0 is the d component of the initial velocity in the Frenet coordinate system
  • alpha and beta are the parameters after model training, which can be trained by linear regression .
  • Step 403 Perform a motion trajectory prediction on the target according to the initial state and the one or more target states to obtain one or more predicted motion trajectories.
  • the probability of each of the multiple predicted motion trajectories can be obtained according to the probability of each target state determined in step 402, or the probability of each of the multiple predicted motion trajectories can be obtained only according to the target state with the highest probability and the initial state
  • the target performs motion trajectory prediction, thereby predicting a motion trajectory with the largest probability.
  • the predicted one or more motion trajectories and their probabilities can be provided to the vehicle path planning module or other control modules, which can provide a reference for subsequent more accurate regulatory decisions.
  • the path planning module can perform necessary avoidance operations according to the predicted trajectory of the target vehicle to prevent collisions with it. For example, for a target vehicle in the adjacent lane on the left or right side of the vehicle’s lane, when the predicted trajectory of the target vehicle invades the vehicle’s lane and conflicts with the path planned by the vehicle at the current moment, the predicted target The trajectory of the vehicle at different times in the future will affect the intrusion position and degree of the intrusion of the planned path of the own vehicle, adjust the planned path and planned speed of the own vehicle, and avoid by reducing the planned speed or offsetting the planned path.
  • Step 403 specifically includes:
  • Step 403a Draw the movement track of the target in the Frenet coordinate system according to the initial state and the target state of the target in the Frenet coordinate system.
  • Cubic Hermite Spline interpolation can be used to fit a smooth curve as the trajectory in the Frenet coordinate system, or the final Frenet coordinates can be obtained through the PID controller (proportional-integral-derivative controller) based on the smooth curve Tie down the track.
  • PID controller proportional-integral-derivative controller
  • Step 404b Convert the one or more predicted motion trajectories in the Frenet coordinate system into a coordinate system for target motion planning control, where the coordinate system for target motion planning control includes a vehicle body coordinate system, Station center coordinate system or world coordinate system.
  • Figure 6 is a schematic diagram of transforming the target's trajectory from the Frenet coordinate system to the station center coordinate system. The dashed line is the predicted trajectory line.
  • the target position is relative to the starting position along the s axis and d of the Frenet coordinate system. All axes have displacements.
  • the displacement transformed to the station center coordinate system is expressed as a moving target extending along the center line of the lane on which the Frenet coordinate system is based. The direction moves forward, one side is closer to the centerline of the lane.
  • the embodiment of the present invention not only uses the initial state of the target, but also uses the preset path information, and uses the motion model to generate the target state in the Frenet coordinate system based on the center line of the lane, and combines the initial state of the target.
  • the state and target state generate predicted trajectories. Compared with the traditional trajectory prediction method based solely on the initial state, it is more in line with the actual motion scene of the moving target, especially the vehicle target, and the predicted trajectory has higher accuracy.
  • FIG. 7 is a structural block diagram of a motion trajectory prediction device provided in Embodiment 2 of the present invention. As shown in Figure 7, the motion trajectory prediction device includes:
  • the initial state acquisition module 71 is configured to acquire the initial state of the target, the initial state including the initial position and the initial motion state;
  • the target state generating module 72 is configured to generate one or more target states of the target according to the initial state and preset path information, and the target state includes at least a target location.
  • the destination state generation module 72 further includes:
  • the associated path determination module 73 is configured to determine one or more preset paths associated with the target according to the initial position and the preset path information;
  • the model calculation module 74 is configured to generate the one or more target states according to the initial state and the motion model, and the one or more target states correspond to the one or more preset paths;
  • the probability calculation module 76 is configured to generate the probability of each target state according to the initial state of the target and the preset path information when the one or more target states are multiple. The probability of is used to obtain the probability of each of the multiple predicted motion trajectories.
  • the model calculation module 74 further includes a coordinate system establishment module 75 for establishing one or more Frenet coordinate systems based on the one or more preset paths.
  • the one or more target states are respectively generated by the model calculation module in the one or more Frenet coordinate systems according to the initial state and the motion model of the target.
  • the motion trajectory prediction device also includes:
  • the motion trajectory prediction module 77 is configured to perform motion trajectory prediction on the target according to the initial state and the one or more target states to obtain one or more predicted motion trajectories;
  • the coordinate system conversion module 78 is configured to convert the one or more predicted motion trajectories into a coordinate system used for target motion planning control.
  • the coordinate system used for target motion planning control includes a vehicle body coordinate system, Station center coordinate system or world coordinate system.
  • the software or firmware includes but is not limited to computer program instructions or codes, and can be executed by a hardware processor.
  • the hardware includes, but is not limited to, various integrated circuits, such as a central processing unit (CPU), a digital signal processor (DSP), a field programmable gate array (FPGA), or an application specific integrated circuit (ASIC).
  • CPU central processing unit
  • DSP digital signal processor
  • FPGA field programmable gate array
  • ASIC application specific integrated circuit
  • Fig. 8 is a structural block diagram of a motion trajectory prediction device provided in Embodiment 3 of the present invention.
  • the motion trajectory prediction device 8 includes a memory 81 and a processor 82.
  • the memory 81 stores computer program instructions, and the processor 82 runs the computer program instructions to perform the motion trajectory prediction related operations described in the first embodiment.
  • the processor 82 is also connected to one or more sensors outside the motion trajectory prediction device 8 to receive the raw data of the surrounding environment of the vehicle detected by the sensors.
  • the sensors include, but are not limited to, cameras, lidar, ultrasonic radar, or millimeter wave radar.
  • the target trajectory prediction result output by the motion trajectory prediction device 8 is generally sent to the path planning and control module of the intelligent driving vehicle to provide reference information for controlling the vehicle.
  • the path planning and control module may also be a software module executed by the processor 82 or integrated in the processor 82, which is not limited in this embodiment.
  • the processor 82 includes, but is not limited to, various types of CPUs, DSPs, microcontrollers, microprocessors, or artificial intelligence processors.
  • the aforementioned motion trajectory prediction devices shown in Figures 7 and 8 introduce preset path information to make the predicted trajectory of the target moving object constrained by the preset path information, which is more in line with the actual motion scene of the moving target, especially the vehicle target.
  • the initial state and motion model of the target are used to generate the target state in the Frenet coordinate system based on the associated preset path, and the predicted trajectory of the target is generated according to the initial state and target state in the Frenet coordinate system, effectively moving the target Combined with the preset path information, the accuracy and reliability of target trajectory prediction are greatly improved.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website, computer, server, or data center.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrated with one or more available media.
  • the usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)).
  • the disclosed system, device and method can be implemented in other ways within the scope of this application.
  • the above-described embodiments are only illustrative.
  • the division of the modules or units is only a logical function division.
  • there may be other division methods for example, multiple units or components may be combined. Or it can be integrated into another system, or some features can be ignored or not implemented.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. .
  • Some or all of the modules may be selected according to actual needs to achieve the objectives of the solutions of the embodiments. Those of ordinary skill in the art can understand and implement it without creative work.
  • the described systems, devices, methods, and schematic diagrams of different embodiments can be combined or integrated with other systems, modules, technologies, or methods without departing from the scope of the present application.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electronic, mechanical or other forms.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Human Computer Interaction (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Optics & Photonics (AREA)
  • Electromagnetism (AREA)
  • Acoustics & Sound (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种运动轨迹预测方法和装置，解决了现有技术中运动轨迹预测精度较低的技术问题,可以应用于包括自动驾驶和辅助驾驶的智能驾驶系统。具体为：获取包括初始位置和初始运动状态的运动目标的初始状态，根据运动目标的初始状态和预设的路径信息生成运动目标的一个或多个目的状态（402），根据运动目标的初始状态和所述一个或多个目的状态对所述运动目标进行运动轨迹预测，得到一个或多个预测的运动轨迹（403）。

Description

运动轨迹预测方法和装置
相关申请的交叉引用
本申请要求在2019年07月27日提交中国国家知识产权局、申请号为201910685546.8、申请名称为“运动轨迹预测方法和装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及运动轨迹预测技术，尤其涉及一种运动轨迹预测方法和装置。
背景技术
智能驾驶系统通常包括感知、定位和规控等模块。感知模块用于感知车辆周围的环境，定位模块用于确定车辆所在的位置，规控模块用于结合感知模块的感知结果和定位模块的定位结果对智能驾驶车辆进行路径规划和控制。智能驾驶车辆在控车的过程中，不仅需要对车辆自身周围环境充分感知，还需要提前预知其未来的轨迹变化，从而避免可能的碰撞。目标轨迹预测往往位于规控模块，是基于对目标物的感知结果对目标物的未来轨迹进行预测，正确的轨迹预测能够为规控模块的路径规划提供重要的参考信息。
现有技术通常基于目标的位置和速度，以一定的运动学模型(如恒定速度、恒定加速度或恒定角速度等)进行轨迹预测。如图1所示，实线表示目标的历史轨迹，虚线表示根据目标当前时刻的位置和速度推理得到的预测轨迹。由于目标未来的运动不会完全按照原始运动特征持续运动，尤其当其未来的运动根据道路结构调整时，采用现有技术进行轨迹预测存在较大偏差。
发明内容
针对现有技术中对运动轨迹预测精度低的技术问题，本发明实施例提出一种运动轨迹预测方法和装置，用于提高运动轨迹预测的精度。
本发明实施例可以应用于包括自动驾驶和辅助驾驶的智能驾驶系统，所述目标不仅可以是车辆目标，还可以是行人、骑行者、机器人或其它运动目标。
第一方面，提供一种运动轨迹预测方法，包括：获取目标的初始状态，所述初始状态包括初始位置和初始运动状态；根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态，所述目的状态至少包括目的位置；根据所述初始状态和所述目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹。通过引入诸如道路结构、路标指示或地图等形式的预设路径信息，使目标的预测轨迹受到预设路径信息的约束，相比单独利用目标的初始状态预测出的轨迹，更符合运动目标尤其是车辆目标的实际运动场景，能够大大改善运动轨迹预测的精度，可以针对长时间(比如10s)的轨迹进行预测。
根据第一方面，在所述运动轨迹预测方法的第一种可能的实现方式中，所述根据所述 初始状态和预设的路径信息生成所述目标的一个或多个目的状态包括：根据所述初始位置和所述预设的路径信息确定与所述目标相关联的一个或多个预设路径；根据所述初始状态和运动模型生成所述一个或多个目的状态，所述一个或多个目的状态对应于所述一个或多个预设路径。确定与所述目标相关联的预设路径，有助于将所述目标的初始状态与所述预设的路径信息关联起来，从而得到与预设的路径信息相关联的目的状态。被确定的与所述目标相关联的预设路径可以是与所述目标距离最近的一个预设路径，也可以是与所述目标临近的多个预设路径。
根据第一方面或第一方面的第一种可能的实现方式，在所述运动轨迹预测方法的第二种可能的实现方式中，所述根据所述初始状态和运动模型生成所述一个或多个目的状态包括：基于所述一个或多个预设路径建立一个或多个Frenet坐标系；分别在所述一个或多个Frenet坐标系中，根据所述目标的初始状态和运动模型生成所述一个或多个目的状态。通过基于与所述目标相关联的预设路径建立Frenet坐标系，便于在与所述目标相关联的预设路径上应用运动模型进行计算。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第三种可能的实现方式中，所述一个或多个预测的运动轨迹为在所述一个或多个Frenet坐标系下的运动轨迹，所述方法还包括：将所述一个或多个预测的的运动轨迹转换到用于目标运动规划控制的坐标系中。将预测的运动轨迹从Frenet坐标系转换到用于目标运动规划控制的坐标系，便于将目标的运动轨迹预测结果应用于运动规划控制模块。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第四种可能的实现方式中，所述用于目标运动规划控制的坐标系包括车体坐标系、站心坐标系或世界坐标系。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第五种可能的实现方式中，所述根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态包括：根据所述初始状态、预设的路径信息和模型训练后的参数生成所述目标的一个或多个目的状态。采用模型训练后的参数生成的目的状态与实际目标运动特征的匹配度更高，轨迹的预测更加精准。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第六种可能的实现方式中，当所述一个或多个目的状态为多个目的状态时，所述方法还包括：根据所述目标的初始状态和所述预设的路径信息生成所述多个目的状态中每个目的状态的概率，所述每个目的状态的概率用于得到多个预测的运动轨迹中每一个的概率。实际运动环境中目标下一步可能的运动轨迹可能有多个，通过引入概率计算，有助于对所述目标可能做出的选择进行更好的判断，进一步提高运动轨迹预测的精准度。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第七种可能的实现方式中，所述根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹包括：根据所述初始状态和具有最大概率的目的状态对所述目标进行运动轨迹预测，得到最大概率的预测的运动轨迹。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第八种可能的实现方式中，所述目的状态还包括目的运动状态，所述目的运动状态包括速度、加速度、角速度或角加速度中至少一个。综合目的位置和目的运动状态进行运动轨迹预测 可以进一步提高预测精准度。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第九种可能的实现方式中，所述初始运动状态包括速度、加速度、角速度或角加速度中至少一个。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第十种可能的实现方式中，所述目标的初始状态与所述预设的路径信息基于相同的坐标系，所述获取目标的初始状态包括：从一种或多种传感器获取目标感知结果；将所述目标感知结果投影到所述相同的坐标系中以得到所述目标的初始状态。传感器输出的结果通常与预设的路径信息基于不同的坐标系，通过投影操作将二者对应到相同坐标系中，便于后续操作中综合二者生成目标的目的状态。
根据第一方面，或以上第一方面的任意一种实现方式，在所述运动轨迹预测方法的第十一种可能的实现方式中，所述目标感知结果基于传感器坐标系或车体坐标系，所述相同的坐标系是站心坐标系或世界坐标系。
第二方面，提供一种运动轨迹预测装置，包括：初始状态获取模块，用于获取目标的初始状态，所述初始状态包括初始位置和初始运动状态；目的状态生成模块，用于根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态，所述目的状态至少包括目的位置；运动轨迹预测模块，用于根据所述初始状态和所述目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹。
根据第二方面，在所述运动轨迹预测装置的第一种可能的实现方式中，所述目的状态生成模块包括：关联路径确定模块，用于根据所述初始位置和所述预设的路径信息确定与所述目标相关联的一个或多个预设路径；模型计算模块，用于根据所述初始状态和运动模型生成所述一个或多个目的状态，所述一个或多个目的状态对应于所述一个或多个预设路径。
根据第二方面或第二方面的第一种可能的实现方式，在所述运动轨迹预测装置的第二种可能的实现方式中，所述模型计算模块包括：坐标系建立模块，用于基于所述一个或多个预设路径建立一个或多个Frenet坐标系；所述一个或多个目的状态由所述模型计算模块分别在所述一个或多个Frenet坐标系中根据所述目标的初始状态和运动模型生成。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第三种可能的实现方式中，所述一个或多个预测的运动轨迹为在所述一个或多个Frenet坐标系下的运动轨迹，所述装置还包括坐标系转换模块，用于将所述一个或多个预测的的运动轨迹转换到用于目标运动规划控制的坐标系中。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第四种可能的实现方式中，所述用于目标运动规划控制的坐标系包括车体坐标系、站心坐标系或世界坐标系。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第五种可能的实现方式中，所述根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态包括：根据所述初始状态、预设的路径信息和模型训练后的参数生成所述目标的一个或多个目的状态。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第 六种可能的实现方式中，所述目的状态生成模块包括概率计算模块，用于当所述一个或多个目的状态为多个目的状态时，根据所述目标的初始状态和所述预设的路径信息生成所述多个目的状态中每个目的状态的概率，所述每个目的状态的概率用于得到多个预测的运动轨迹中每一个的概率。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第七种可能的实现方式中，所述根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹包括：根据所述初始状态和具有最大概率的目的状态对所述目标进行运动轨迹预测，得到最大概率的预测的运动轨迹。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第八种可能的实现方式中，所述目的状态还包括目的运动状态，所述目的运动状态包括速度、加速度、角速度或角加速度中至少一个。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第九种可能的实现方式中，所述初始运动状态包括速度、加速度、角速度或角加速度中至少一个。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第十种可能的实现方式中，所述所述目标的初始状态与所述预设的路径信息基于相同的坐标系，所述获取目标的初始状态包括：从一种或多种传感器获取目标感知结果；将所述目标感知结果投影到所述相同的坐标系中以得到所述目标的初始状态。传感器输出的结果通常与预设的路径信息基于不同的坐标系，通过投影操作将二者对应到相同坐标系中，便于后续操作中综合二者生成目标的目的状态。
根据第二方面，或以上第二方面的任意一种实现方式，在所述运动轨迹预测装置的第十一种可能的实现方式中，所述目标感知结果基于传感器坐标系或车体坐标系，所述相同的坐标系是站心坐标系或世界坐标系。
第三方面，提供一种运动轨迹预测装置，包括存储器和处理器，所述存储器存储计算机程序指令，所述处理器运行所述计算机程序指令以执行权利要求1-8任一项所述的方法。
第四方面，提供一种计算机存储介质，包括计算机指令，当所述计算机指令在被处理器运行时，使得所述运动轨迹预测装置执行如权利要求1-8任一项所述的方法。
第五方面，提供一种计算机程序产品，当所述计算机程序产品在处理器上运行时，使得所述运动轨迹预测装置执行如权利要求1-8任一项所述的方法。
附图说明
图1为现有技术运动轨迹预测方法的示意图；
图2为本发明实施例中Frenet坐标系的的示意图；
图3为本发明实施例典型的应用场景图(以机动车在道路上的行驶情景为例)；
图4为本发明实施例一提供的运动轨迹预测方法的流程图；
图5为本发明实施例一中一个目标对应多个预设路径的情况示意图；
图6为本发明实施例一中将目标的运动轨迹由Frenet坐标系转换到站心坐标系的示意图；
图7为本发明实施例二提供的运动轨迹预测装置的结构框图；
图8为本发明实施例三提供的运动轨迹预测装置的结构框图。
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，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
无论是高级辅助驾驶ADAS系统，还是自动驾驶系统，一个重要工作就是计算系统自己的位置，以及系统自己与道路、车辆或行人等交通元素之间的相对位置关系或速度关系。为了描述这些复杂的空间关系，需要建立空间坐标系，这是实现各种环境感知和决策规划算法的前提条件。首先对本发明实施例涉及的几种坐标系作如下说明：
(1)传感器坐标系：各传感器感测结果所基于的坐标系，根据传感器种类或安装位置的不同，各传感器所采用的坐标系也可能不同。
(2)车体坐标系：用来描述车辆周围的物体和本车之间的相对位置关系，三个相互垂直的坐标轴分别为汽车的长度方向、宽度方向和高度方向。目前学术界和工业界有几种比较常用的车体坐标系定义方式，分别是ISO国际标准定义、SAE(Society of Automotive Engineers)汽车工程师协会定义和基于惯用测量单元IMU的坐标定义。
(3)站心坐标系：又称ENU坐标系或NEU坐标系或东北天坐标系，三个坐标轴分别指向相互垂直的东向、北向和天向。
(4)世界坐标系：世界坐标系是描述地球上位置关系的坐标系，常见的世界坐标系包括WGS-84坐标系、UTM(Universal Transverse Mercartor，通用横轴横墨卡托)坐标系。
(5)Frenet坐标系:如图2所示，使用预设路径的中心线作为参考线，定义坐标系的原点为参考线上的某一参考点，坐标系的s轴为运动目标沿所述参考线的前进方向，坐标系的d轴为所述运动目标在参考线上的投影点与所述目标之间的连线方向，所述运动目标与所述投影点之间的连线垂直于所述参考线在所述投影点的切线。
图3以机动车在道路上的行驶情景为例介绍本发明实施例典型的应用场景图，但本发明实施例的应用场景不仅仅限于这一具体情形，也可应用于所有运动目标(包括机动车、非机动车、行人或机器人等)在所有形式的路径(包括高速公路、城市道路、乡村道路或室内路径等)上的运动，后续实施例以车辆或车道为例做描述，但本领域技术人员可以将其扩展至其他目标的运动轨迹预测领域。如图3所示，在t 0时刻，目标车辆在位置A以速度V 0(本发明实施例中的速度包括速度大小和方向)沿着车道中心线右侧行驶，前方不远处车道将由一条分为两条(分别是直行和右转)，因此在待预测的t时刻该目标车辆较大可能地出现在位置B或位置C，从而目标车辆在t 0时刻至t时刻的运动轨迹存在两种并列的可能，即继续直行(由位置A行驶至位置B)或者右转(由位置A行驶至位置C)。
图4是本发明运动轨迹预测方法的一个实施例的流程图。如图4所示，运动轨迹预测方法包括步骤401、步骤402和步骤403：
步骤401，获取目标的初始状态。所述目标包括行人、机动车或非机动车，也可以包 括其它运动目标。通过一种或者多种目标探测传感器(如摄像头、激光雷达或毫米波雷达等)获取环境探测的原始数据，再通过一定的目标检测和跟踪算法获取目标感知结果，包括位置、速度、加速度或角速度等运动状态相关信息。由于目标探测传感器获取的目标感知结果所基于的坐标系(常见的例如传感器坐标系或车体坐标系)经常不同于预设的路径信息所基于的坐标系(常见的例如站心坐标系或世界坐标系)，因此步骤401包括坐标转换操作，将所述目标感知结果投影到所述预设的路径信息所基于的坐标中，得到基于与所述预设的路径信息相同的坐标系的目标的初始状态。所述目标的初始状态包括初始位置和初始运动状态，所述初始运动状态包括速度、加速度、角速度或角加速度。
步骤402，根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态。所述预设的路径信息可以是道路结构、路标指示或地图等形式的信息。所述目的状态至少包括目的位置，还可以包括目的运动状态，如速度、加速度、角速度或角加速度。综合目的位置和目的运动状态进行运动轨迹预测可以进一步提高预测精准度。
步骤402包括：根据目标初始状态中的初始位置信息和所述预设的路径信息确定与所述目标相关联的一个或多个预设路径。具体地，以图3所示情形为例，可以根据目标车辆在t 0时刻的初始位置查找道路交通图等形式的预设的路径信息，获取目标车辆在t 0时刻所对应的车道及车道中心线，该车道及车道中心线为图3中与目标车辆相关的预设路径。
需要注意的是，由于目标一般为具有一定形状的物体，因此可以一个目标对应多个车道，图5是一个目标对应多个预设路径的情况示意图，与图3所示的以机动车在道路上的行驶情景为例的应用场景相对应。起始点A是目标车辆的初始位置，速度V 0是目标车辆的初始运动状态，二者包括在目标车辆的初始状态中。由于在初始时刻t 0目标车辆位于第一车道和第二车道相重合的车道内，因此可将第一车道中心线和第二车道中心线作为与目标车辆相关联的多个预设路径。在其他场景下，可以选择与运动目标初始位置距离最近的一个预设路径作为关联预设路径，也可以选择与运动目标的距离在设定阈值内的多个预设路径作为所述相关联的预设路径，或者选择满足设定数量的与运动目标距离最近的多个预设路径作为所述相关联的预设路径，还可以选择与运动目标的运动方向(如速度方向、加速度方向或角加速度方向)最匹配的一个或多个预设路径作为所述相关联的预设路径。当选择与所述目标相关联的多个预设路径时，可通过计算每个相关联的预设路径的概率得到与所述每个相关联的预设路径相对应的目的状态的概率，所述目的状态的概率用于计算预测的运动轨迹的概率，可为后续作出更准确的规控决策提供参考。
步骤402还包括：基于选择的一个或多个预设路径建立一个或多个Frenet坐标系，分别在所述一个或多个Frenet坐标系中，根据所述目标的初始状态和运动模型生成所述一个或多个目的状态。
以图5所示一个目标对应多个关联的预设路径的情况为例，将目标初始位置起始点A分别投影到第一车道中心线和第二车道中心线，分别获取初始位置投影点1和投影点2。分别以投影点1和投影点2为原点，以对应的第一车道中心线和第二车道中心线的方向为s方向，沿车道中心线建立两个Frenet坐标系，则起始点A在两个Frenet坐标系中的s分量均为0，d分量分别由起始点A与投影点1、投影点2各自的距离确定。分别在两个Frenet坐标系中，根据所述目标车辆的初始状态和运动模型生成两个目的状态，两个目的状态分别对应于之前选择的两条关联的预设路径(即第一车道中心线和第二车道中心 线)，对应于第一车道中心线的目的状态为目的点B和速度V t，对应于第二车道中心线的目的状态为目的点C和速度V t’。
将目标的初始位置和初始速度投影到同一Frenet坐标系中，如前所述，Frenet坐标系的s轴为目标沿关联预设路径(例如图5的第一车道中心线或第二车道中心线)的前进方向，坐标系的d轴为所述目标在关联预设路径上的投影点与所述目标之间的连线方向，所述目标与所述投影点之间的连线垂直于所述关联预设路径在所述投影点的切线，以所述投影点作为原点，0时刻目标的初始位置为P 0(P 0在s轴和d轴的分量分别为P 0S和P 0d)，目标的初始速度为V 0(V 0在s轴和d轴的分量分别为V 0S和V 0d)，待预测的T时刻目标的目的位置为P T(P T在s轴和d轴的分量分别为P TS和P Td)，目标的目的速度为V T(V T在s轴和d轴的分量分别为V TS和V Td)。下面用两个例子说明如何利用初始状态和运动模型计算目的状态：
(1)如果设置在s轴和d轴上采用的运动模型分别为匀速运动模型和匀减速运动模型，并假设T时刻目标的速度方向平行于关联的预设路径，则待预测的T时刻的目的状态为：
由于目标在s轴上保持恒定速度V 0S，所以P TS＝V 0S×T；
由于目标在d轴上匀减速，因此P Td＝P 0d-0.5×V 0d×T；
由于目标在s轴上保持恒定速度V 0S，所以V TS＝V 0S；
由于T时刻目标的速度方向平行于关联的预设路径，而d轴与关联的预设路径相垂直，因此V Td＝0。
(2)如果设置在s轴和d轴上采用的运动模型分别为匀加速运动模型和匀速运动模型，并假设目标的加速度为a，则待预测的T时刻的目的状态为：
P TS＝0.5×V 0S×T+0.5×aT 2；
P Td＝P 0d-V 0d×T；
V TS＝V 0S+aT；
V Td＝V 0d。
由前面可知，一个目标可能关联多个预设路径，对应多个Frenet坐标系，也会生成多个对应的目的状态，此时还要根据所述目标的初始状态生成每个目的状态的概率，针对说明书附图5所示情形，可以分别生成目的位置B和目的位置C的概率，将两个概率连同两个目的状态一同提供给路径规划与控制模块，以更好地辅助规控决策；或者可以上述两个概率确定概率最大的一个目的状态，仅将所述概率最大的一个目的状态提供给路径规划与控制模块。
为了进一步提高轨迹预测的精度，步骤402“根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态”时还采用了模型训练后的参数。模型训练是指使用已有的样本数据，通过一些方法确定模型的参数。下面通过三个例子进行说明本实施例如何使用模型训练后的参数：
(1)采用模型训练后的参数选择与目标相关联的一个或多个预设路径。
对于与目标相关联的车道的选择，可以结合目标初始位置与各个车道的距离以及初始速度方向与车道线的夹角建立多分类器模型，通过softmax、boosting或决策树等算法获取对应于各个车道的概率。以softmax为例，根据目标初始位置与各个车道的距离以及初 始速度方向与车道线的夹角建立如下模型：
w＝alpha1×d 0+alpha2×Δh 0
p j＝exp(-w j)/(∑ iexp(-w i))
其中w为用于计算目标关联于对应车道的概率的权重，d 0为目标初始位置与车道的距离(即目标初始位置到车道中心线的投影与所述目标初始位置之间的距离)，Δh 0为目标初始速度方向与车道中心线朝向的夹角，alpha1和alpha2为模型训练后的参数，p j为目标关联于车道j的概率，j为整数，w i为目标关联于车道i的概率的权重，i需要遍历所有车道。
(2)采用模型训练后的参数计算目标的目的位置。
以目的位置在Frenet坐标系的d分量为例，可以构建如下位移衰减模型
d f＝exp(-alpha×l-beta)(d 0)
其中l为预测轨迹的长度，由目的位置在Frenet坐标系的s分量决定，d 0为初始位置在Frenet坐标系的d分量，alpha和beta为模型训练后的参数，可以通过线性回归来训练参数。
(3)采用模型训练后的参数计算目标的目的速度。
以目的速度在Frenet坐标系的d分量为例，可以构建如下位移衰减模型
v f＝exp(-alpha×l-beta)(v 0)
其中l为预测轨迹的长度，由目的位置在Frenet坐标系的s分量决定，v 0为初始速度在Frenet坐标系的d分量，alpha和beta为模型训练后的参数，可以通过线性回归来训练参数。
步骤403，根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹。当目的状态有多个时，可以根据步骤402确定的每个目的状态的概率得到多个预测的运动轨迹中每一个的概率，也可以仅根据概率最大的一个目的状态结合所述初始状态对所述目标进行运动轨迹预测，从而预测概率最大的一个运动轨迹。预测的一个或多个运动轨迹及其概率可提供给车辆路径规划模块或者其他的控制模块，可为后续作出更准确的规控决策提供参考。具体来看，路径规划模块可以根据目标车辆的预测轨迹进行必要的避让操作，防止与其发生碰撞。例如，对于在自车车道左侧或者右侧相邻车道上的目标车辆，当对目标车辆预测的轨迹侵占自车的车道、与当前时刻自车规划的路径有冲突时，则根据预测的目标车辆在未来不同时刻的运动轨迹对自车规划的路径的入侵位置和入侵程度，对自车的规划路径和规划速度进行调整，通过降低规划速度或者偏移规划路径进行避让。
步骤403具体包括：
步骤403a，根据目标在Frenet坐标系的初始状态和目的状态绘制目标在Frenet坐标系的运动轨迹。具体地，可以采用Cubic Hermite Spline插值的方式拟合一条平滑曲线作为Frenet坐标系下的轨迹，也可以以该平滑曲线为基础通过PID控制器(比例-积分-微分控制器)获取最终的Frenet坐标系下轨迹。
步骤404b，将Frenet坐标系下的所述一个或多个预测的的运动轨迹转换到用于目标运动规划控制的坐标系中，所述用于目标运动规划控制的坐标系包括车体坐标系、站心坐标系或世界坐标系。图6是将目标的运动轨迹由Frenet坐标系转换到站心坐标系的示意图，其中的虚线为预测的轨迹线，在Frenet坐标系下目的位置相对起始位置沿Frenet坐标系的s轴和d轴均有位移，结合Frenet坐标系中s轴和d轴的意义可知，所述位移变换到在站心坐标系就表现为运动目标一边沿着所述Frenet坐标系所基于的车道中心线伸 展的方向向前运动，一边更靠近车道中心线。
从本实施例可以看出，本发明实施例不仅利用了目标物的初始状态，还利用了预设的路径信息，采用运动模型在基于车道中心线的Frenet坐标系生成目的状态，并结合目标初始状态和目的状态生成预测轨迹，相比传统的单纯基于初始状态的轨迹预测方法，更符合运动目标尤其是车辆目标的实际运动场景，预测的轨迹具有更高的准确性。
图7为本发明实施例二提供的运动轨迹预测装置的结构框图。如图7所示，运动轨迹预测装置包括：
初始状态获取模块71，用于获取目标的初始状态，所述初始状态包括初始位置和初始运动状态；
目的状态生成模块72，用于根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态，所述目的状态至少包括目的位置。
该目的状态生成模块72进一步包括：
关联路径确定模块73，用于根据所述初始位置和所述预设的路径信息确定与所述目标相关联的一个或多个预设路径；
模型计算模块74，用于根据所述初始状态和运动模型生成所述一个或多个目的状态，所述一个或多个目的状态对应于所述一个或多个预设路径；
概率计算模块76，用于当所述一个或多个目的状态为多个时，根据所述目标的初始状态和所述预设的路径信息生成每个目的状态的概率，所述每个目的状态的概率用于得到多个预测的运动轨迹中每一个的概率。
其中，模型计算模块74又进一步包括坐标系建立模块75，用于基于所述一个或多个预设路径建立一个或多个Frenet坐标系。所述一个或多个目的状态由所述模型计算模块分别在所述一个或多个Frenet坐标系中根据所述目标的初始状态和运动模型生成。
运动轨迹预测装置还包括：
运动轨迹预测模块77，用于根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹；
坐标系转换模块78，用于将所述一个或多个预测的的运动轨迹转换到用于目标运动规划控制的坐标系中，所述用于目标运动规划控制的坐标系包括车体坐标系、站心坐标系或世界坐标系。
图7中的各个模块的只一个或多个可以软件、硬件、固件或其结合实现。所述软件或固件包括但不限于计算机程序指令或代码，并可以被硬件处理器所执行。所述硬件包括但不限于各类集成电路，如中央处理单元(CPU)、数字信号处理器(DSP)、现场可编程门阵列(FPGA)或专用集成电路(ASIC)。
图8是本发明实施例三提供的运动轨迹预测装置的结构框图。如图8所示，运动轨迹预测装置8包括存储器81和处理器82。所述存储器81存储计算机程序指令，所述处理器82运行所述计算机程序指令以执行实施例一所描述的运动轨迹预测相关操作。所述处理器82还与运动轨迹预测装置8外界的一个或多个传感器相连接，接收所述传感器探测的自车周围环境的原始数据。所述传感器包括但不限于如摄像头、激光雷达、超声波雷达或毫米波雷达。运动轨迹预测装置8输出的目标轨迹预测结果一般发送给智能驾驶车辆的路径规划与控制模块，提供控车参考信息。路径规划与控制模块也可以是由处理器82执 行的一个软件模块或集成于处理器82中，本实施例不做限定。处理器82包括但不限于各类CPU、DSP、微控制器、微处理器或人工智能处理器。
上述图7、图8所示的运动轨迹预测装置，通过引入预设路径信息，使目标运动物体的预测轨迹受到预设路径信息的约束，更符合运动目标尤其是车辆目标的实际运动场景。另外，利用目标的初始状态和运动模型在以关联的预设路径为基础的Frenet坐标系生成目的状态，并依据Frenet坐标系下的初始状态和目的状态生成目标的预测轨迹，有效地将目标运动和预设路径信息结合起来，大大改善目标轨迹预测的精度和可靠性。
本领域的技术人员可以清楚地了解到，本申请提供的各实施例的描述可以相互参照，为描述的方便和简洁，例如关于本申请实施例提供的各装置、设备的功能以及执行的步骤可以参照本申请方法实施例的相关描述，各方法实施例之间、各装置实施例之间也可以互相参照。
本领域技术人员可以理解：实现上述各方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成。前述的程序可以存储于一计算机可读取存储介质中。该程序在执行时，执行包括上述各方法实施例的全部或部分步骤；而前述的存储介质包括：ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
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Solid State Disk(SSD))等。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，在没有超过本申请的范围内，可以通过其他的方式实现。例如，以上所描述的实施例仅仅是示意性的，例如，所述模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性劳动的情况下，即可以理解并实施。
另外，所描述系统、装置和方法以及不同实施例的示意图，在不超出本申请的范围内，可以与其它系统，模块，技术或方法结合或集成。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电子、机械或其它的形式。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应所述以权利要求的保护范围为准。

Claims (23)

  1. 一种运动轨迹预测方法，其特征在于，包括：
    获取目标的初始状态，所述初始状态包括初始位置和初始运动状态；
    根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态，所述目的状态至少包括目的位置；
    根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态包括：
    根据所述初始位置和所述预设的路径信息确定与所述目标相关联的一个或多个预设路径；
    根据所述初始状态和运动模型生成所述一个或多个目的状态，所述一个或多个目的状态对应于所述一个或多个预设路径。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述初始状态和运动模型生成所述一个或多个目的状态包括：
    基于所述一个或多个预设路径建立一个或多个Frenet坐标系；
    分别在所述一个或多个Frenet坐标系中，根据所述目标的初始状态和运动模型生成所述一个或多个目的状态。
  4. 根据权利要求3所述的方法，其特征在于，所述一个或多个预测的运动轨迹为在所述一个或多个Frenet坐标系下的运动轨迹，所述方法还包括：
    将所述一个或多个预测的的运动轨迹转换到用于目标运动规划控制的坐标系中。
  5. 根据权利要求4所述的方法，其特征在于，所述用于目标运动规划控制的坐标系包括车体坐标系、站心坐标系或世界坐标系。
  6. 根据权利要求1-5任一项所述的方法，其特征在于，所述根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态包括：根据所述初始状态、预设的路径信息和模型训练后的参数生成所述目标的一个或多个目的状态。
  7. 根据权利要求1-6任一项所述的方法，其特征在于，还包括：当所述一个或多个目的状态为多个目的状态时，所述方法还包括：
    根据所述目标的初始状态和所述预设的路径信息生成所述多个目的状态中每个目的状态的概率，所述每个目的状态的概率用于得到多个预测的运动轨迹中每一个的概率。
  8. 根据权利要求7所述的方法，其特征在于，所述根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹包括：根据所述初始状态和具有最大概率的目的状态对所述目标进行运动轨迹预测，得到最大概率的预测的运动轨迹。
  9. 根据权利要求1-8任一项所述的方法，其特征在于，所述目的状态还包括目的运动状态，所述目的运动状态包括速度、加速度、角速度或角加速度中至少一个。
  10. 根据权利要求1-9任一项所述的方法，其特征在于，所述初始运动状态包括速度、加速度、角速度或角加速度中至少一个。
  11. 一种运动轨迹预测装置，其特征在于，包括：
    初始状态获取模块，用于获取目标的初始状态，所述初始状态包括初始位置和初始运动状态；
    目的状态生成模块，用于根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态，所述目的状态至少包括目的位置；
    运动轨迹预测模块，用于根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹。
  12. 根据权利要求11所述的装置，其特征在于，所述目的状态生成模块包括：
    关联路径确定模块，用于根据所述初始位置和所述预设的路径信息确定与所述目标相关联的一个或多个预设路径；
    模型计算模块，用于根据所述初始状态和运动模型生成所述一个或多个目的状态，所述一个或多个目的状态对应于所述一个或多个预设路径。
  13. 根据权利要求12所述的装置，其特征在于，所述模型计算模块包括：
    坐标系建立模块，用于基于所述一个或多个预设路径建立一个或多个Frenet坐标系；
    所述一个或多个目的状态由所述模型计算模块分别在所述一个或多个Frenet坐标系中根据所述目标的初始状态和运动模型生成。
  14. 根据权利要求13所述的装置，其特征在于，所述一个或多个预测的运动轨迹为在所述一个或多个Frenet坐标系下的运动轨迹，所述装置还包括：
    坐标系转换模块，用于将所述一个或多个预测的的运动轨迹转换到用于目标运动规划控制的坐标系中。
  15. 根据权利要求14所述的装置，其特征在于，所述用于目标运动规划控制的坐标系包括车体坐标系、站心坐标系或世界坐标系。
  16. 根据权利要求11-15任一项所述的装置，其特征在于，所述根据所述初始状态和预设的路径信息生成所述目标的一个或多个目的状态包括：根据所述初始状态、预设的路径信息和模型训练后的参数生成所述目标的一个或多个目的状态。
  17. 根据权利要求11-16任一项所述的装置，其特征在于，所述目的状态生成模块包括概率计算模块，用于当所述一个或多个目的状态为多个目的状态时，根据所述目标的初始状态和所述预设的路径信息生成所述多个目的状态中每个目的状态的概率，所述每个目的状态的概率用于得到多个预测的运动轨迹中每一个的概率。
  18. 根据权利要求17所述的方法，其特征在于，所述根据所述初始状态和所述一个或多个目的状态对所述目标进行运动轨迹预测，得到一个或多个预测的运动轨迹包括：根据所述初始状态和具有最大概率的目的状态对所述目标进行运动轨迹预测，得到最大概率的预测的运动轨迹。
  19. 根据权利要求11-18任一项所述的装置，其特征在于，所述目的状态还包括目的运动状态，所述目的运动状态包括速度、加速度、角速度或角加速度中至少一个。
  20. 根据权利要求11-19任一项所述的装置，其特征在于，所述初始运动状态包括速度、加速度、角速度或角加速度中至少一个。
  21. 一种运动轨迹预测装置，其特征在于，包括：包括存储器和处理器，所述存储器存储计算机程序指令，所述处理器运行所述计算机程序指令以执行权利要求1-10任一项 所述的方法。
  22. 一种计算机存储介质，其特征在于，包括计算机指令，当所述计算机指令在被处理器运行时，使得所述运动轨迹预测装置执行如权利要求1-10任一项所述的方法。
  23. 一种计算机程序产品，其特征在于，当所述计算机程序产品在处理器上运行时，使得所述运动轨迹预测装置执行如权利要求1-10任一项所述的方法。
PCT/CN2020/081894 2019-07-27 2020-03-28 运动轨迹预测方法和装置 WO2021017504A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP20846548.4A EP3992732A4 (en) 2019-07-27 2020-03-28 METHOD AND DEVICE FOR PREDICTING MOVEMENT PATH
US17/585,802 US20220144265A1 (en) 2019-07-27 2022-01-27 Moving Track Prediction Method and Apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910685546.8 2019-07-27
CN201910685546.8A CN112286049A (zh) 2019-07-27 2019-07-27 运动轨迹预测方法和装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/585,802 Continuation US20220144265A1 (en) 2019-07-27 2022-01-27 Moving Track Prediction Method and Apparatus

Publications (1)

Publication Number Publication Date
WO2021017504A1 true WO2021017504A1 (zh) 2021-02-04

Family

ID=74230167

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/081894 WO2021017504A1 (zh) 2019-07-27 2020-03-28 运动轨迹预测方法和装置

Country Status (4)

Country Link
US (1) US20220144265A1 (zh)
EP (1) EP3992732A4 (zh)
CN (1) CN112286049A (zh)
WO (1) WO2021017504A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114872735A (zh) * 2022-07-10 2022-08-09 成都工业职业技术学院 基于神经网络算法的自动驾驶物流车辆决策方法和装置
EP4147936A1 (en) * 2021-12-23 2023-03-15 Baidu USA LLC Drive with caution under uncertainty for an autonomous driving vehicle

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11754666B2 (en) * 2021-02-12 2023-09-12 Aptiv Technologies Limited Radar detection prioritization based on downstream feedback
CN114419098A (zh) * 2022-01-18 2022-04-29 长沙慧联智能科技有限公司 基于视觉变换的运动目标轨迹预测方法及装置
CN114964292B (zh) * 2022-05-30 2023-10-20 国汽智控(北京)科技有限公司 全局路径规划方法、装置、电子设备及存储介质
CN117874529B (zh) * 2024-03-12 2024-05-28 浪潮电子信息产业股份有限公司 运动轨迹预测方法、模型训练方法、装置、设备及介质

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014206654A1 (de) * 2013-06-27 2014-12-31 Bayerische Motoren Werke Aktiengesellschaft Vorhersage von fahrpfaden eines fahrzeugs
US9248834B1 (en) * 2014-10-02 2016-02-02 Google Inc. Predicting trajectories of objects based on contextual information
CN105549597A (zh) * 2016-02-04 2016-05-04 同济大学 一种基于环境不确定性的无人车动态路径规划方法
CN106774327A (zh) * 2016-12-23 2017-05-31 中新智擎有限公司 一种机器人路径规划方法及装置
CN106864361A (zh) * 2017-02-14 2017-06-20 驭势科技（北京）有限公司 车辆与车外人车交互的方法、系统、装置和存储介质
CN107168342A (zh) * 2017-07-12 2017-09-15 哈尔滨工大智慧工厂有限公司 一种用于机器人路径规划的行人轨迹预测方法
US20180074505A1 (en) * 2016-09-14 2018-03-15 Qualcomm Incorporated Motion planning and intention prediction for autonomous driving in highway scenarios via graphical model-based factorization
CN109059944A (zh) * 2018-06-06 2018-12-21 上海国际汽车城（集团）有限公司 基于驾驶习惯学习的运动规划方法
CN109885066A (zh) * 2019-03-26 2019-06-14 北京经纬恒润科技有限公司 一种运动轨迹预测方法及装置

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130238181A1 (en) * 2012-03-12 2013-09-12 Toyota Motor Eng. & Man. North America (Tema) On-board vehicle path prediction using processed sensor information
EP3001272B1 (en) * 2014-09-26 2017-04-12 Volvo Car Corporation Method of trajectory planning for yielding manoeuvres
CN104808679B (zh) * 2015-02-27 2017-06-23 中国民航大学 基于飞行轨迹预测的通用航空aip文件智能匹配方法
US11235777B2 (en) * 2015-10-15 2022-02-01 Harman International Industries, Incorporated Vehicle path prediction and target classification for autonomous vehicle operation
US10029682B2 (en) * 2016-01-22 2018-07-24 Toyota Motor Engineering & Manufacturing North America, Inc. Surrounding vehicle classification and path prediction
DE102017206987A1 (de) * 2017-04-26 2018-10-31 Bayerische Motoren Werke Aktiengesellschaft Verfahren, Computerprogrammprodukt, Computer-lesbares Medium, Steuergerät und Fahrzeug umfassen das Steuergerät zum Bestimmen eines kollektiven Manövers von wenigstens zwei Fahrzeugen
CA3027627C (en) * 2017-07-13 2021-08-10 Beijing Didi Infinity Technology And Development Co., Ltd. Systems and methods for trajectory determination
FR3072069B1 (fr) * 2017-10-10 2019-09-20 Commissariat A L'energie Atomique Et Aux Energies Alternatives Procede de conduite automatique sous contrainte d'un vehicule, notamment d'un bus dans un centre de remisage, et dispositif mettant en œuvre un tel procede
EP3495219B1 (en) * 2017-12-11 2023-07-05 Volvo Car Corporation Path prediction for a vehicle
CN109583151B (zh) * 2019-02-20 2023-07-21 阿波罗智能技术(北京)有限公司 车辆的行驶轨迹预测方法及装置
CN109976355B (zh) * 2019-04-26 2021-12-10 腾讯科技（深圳）有限公司 轨迹规划方法、系统、设备及存储介质

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014206654A1 (de) * 2013-06-27 2014-12-31 Bayerische Motoren Werke Aktiengesellschaft Vorhersage von fahrpfaden eines fahrzeugs
US9248834B1 (en) * 2014-10-02 2016-02-02 Google Inc. Predicting trajectories of objects based on contextual information
CN105549597A (zh) * 2016-02-04 2016-05-04 同济大学 一种基于环境不确定性的无人车动态路径规划方法
US20180074505A1 (en) * 2016-09-14 2018-03-15 Qualcomm Incorporated Motion planning and intention prediction for autonomous driving in highway scenarios via graphical model-based factorization
CN106774327A (zh) * 2016-12-23 2017-05-31 中新智擎有限公司 一种机器人路径规划方法及装置
CN106864361A (zh) * 2017-02-14 2017-06-20 驭势科技（北京）有限公司 车辆与车外人车交互的方法、系统、装置和存储介质
CN107168342A (zh) * 2017-07-12 2017-09-15 哈尔滨工大智慧工厂有限公司 一种用于机器人路径规划的行人轨迹预测方法
CN109059944A (zh) * 2018-06-06 2018-12-21 上海国际汽车城（集团）有限公司 基于驾驶习惯学习的运动规划方法
CN109885066A (zh) * 2019-03-26 2019-06-14 北京经纬恒润科技有限公司 一种运动轨迹预测方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3992732A4 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4147936A1 (en) * 2021-12-23 2023-03-15 Baidu USA LLC Drive with caution under uncertainty for an autonomous driving vehicle
CN114872735A (zh) * 2022-07-10 2022-08-09 成都工业职业技术学院 基于神经网络算法的自动驾驶物流车辆决策方法和装置

Also Published As

Publication number Publication date
CN112286049A (zh) 2021-01-29
EP3992732A1 (en) 2022-05-04
US20220144265A1 (en) 2022-05-12
EP3992732A4 (en) 2022-08-24

Similar Documents

Publication Publication Date Title
WO2021017504A1 (zh) 运动轨迹预测方法和装置
US10591926B2 (en) Smooth road reference for autonomous driving vehicles based on 2D constrained smoothing spline
US10606277B2 (en) Speed optimization based on constrained smoothing spline for autonomous driving vehicles
US10816990B2 (en) Non-blocking boundary for autonomous vehicle planning
US20230080764A1 (en) Deep learning for object detection using pillars
US20190086925A1 (en) Path optimization based on constrained smoothing spline for autonomous driving vehicles
WO2020147311A1 (zh) 车辆行驶保障方法、装置、设备及可读存储介质
US10800427B2 (en) Systems and methods for a vehicle controller robust to time delays
JP2022160538A (ja) 衝突検出方法、装置、電子機器、記憶媒体、自動運転車両及びコンピュータプログラム
US20230205216A1 (en) Ground plane estimation using lidar semantic network
JP2022532920A (ja) レーダーデータからのヨーレート
JP2014025925A (ja) 車両用コントローラ、車両システム
US20220101155A1 (en) Trajectory Generation Using Road Network Model
US20220289179A1 (en) Trajectory checker
US11385642B2 (en) Perpendicular cut-in training
US10796571B2 (en) Method and device for detecting emergency vehicles in real time and planning driving routes to cope with situations to be expected to be occurred by the emergency vehicles
US11367289B1 (en) Machine learning-based framework for drivable surface annotation
WO2022142592A1 (zh) 一种车头先入的泊车方法、装置和系统
US11999372B2 (en) Operation of an autonomous vehicle based on availability of navigational information
GB2604425A (en) Controlling an autonomous vehicle using a proximity rule
WO2023092451A1 (zh) 预测可行驶车道的方法和装置
US20230333237A1 (en) Merge-split techniques for sensor data filtering
CN115127576A (zh) 路径规划方法、装置、芯片、终端、电子设备及存储介质
RU2721436C1 (ru) Способ помощи вождению и устройство помощи вождению
WO2022034815A1 (en) Vehicle surroundings recognition device

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: 20846548

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: 2020846548

Country of ref document: EP

Effective date: 20220128