WO2023230740A1 - 一种异常驾驶行为识别的方法、装置和交通工具 - Google Patents

一种异常驾驶行为识别的方法、装置和交通工具 Download PDF

Info

Publication number
WO2023230740A1
WO2023230740A1 PCT/CN2022/095804 CN2022095804W WO2023230740A1 WO 2023230740 A1 WO2023230740 A1 WO 2023230740A1 CN 2022095804 W CN2022095804 W CN 2022095804W WO 2023230740 A1 WO2023230740 A1 WO 2023230740A1
Authority
WO
WIPO (PCT)
Prior art keywords
parameter
vehicle
manipulation
driving behavior
expected
Prior art date
Application number
PCT/CN2022/095804
Other languages
English (en)
French (fr)
Inventor
胡昊锐
刘利梁
胡峰伟
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/CN2022/095804 priority Critical patent/WO2023230740A1/zh
Publication of WO2023230740A1 publication Critical patent/WO2023230740A1/zh

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
    • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
    • B60W40/08—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to drivers or passengers
    • B60W40/09—Driving style or behaviour
    • 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
    • B60W50/08—Interaction between the driver and the control system
    • B60W50/14—Means for informing the driver, warning the driver or prompting a driver intervention

Definitions

  • Embodiments of the present application relate to the field of safety, and more specifically, to a method, device and vehicle for identifying abnormal driving behavior.
  • the detection of abnormal driving behavior usually relies on cameras to collect images of the user's body or face, and then analyze the user's abnormal driving behavior after feature extraction.
  • This method cannot detect when the light is poor or the user is blocked; there are also methods based on
  • the detection method of wearable sensors uses the user's physiological characteristic data to analyze and identify the user's abnormal driving state. This method is also limited by the collection of user physical data. Moreover, the above methods often require additional sensor costs. Therefore, how to independently and accurately detect users' abnormal driving behavior has become an urgent problem to be solved.
  • Embodiments of the present application provide a method and device for identifying abnormal driving behavior, which can independently and accurately identify the user's abnormal driving behavior without adding additional sensors, and can reduce the cost of detecting abnormal driving behavior.
  • a method for identifying abnormal driving behavior including: obtaining a first manipulation parameter, a first motion state parameter, and first environmental information.
  • the first manipulation parameter is the manipulation that controls the movement of the vehicle at the first moment.
  • Parameters, the first motion state parameter is used to indicate the motion state of the vehicle at the first moment
  • the first environment information is used to indicate the surrounding environment of the vehicle at the first moment; according to the first motion state Parameters and the first environment information, based on the driver model, determine the first expected manipulation parameters; based on the first manipulation parameters and the first expected manipulation parameters, determine whether there is abnormal driving behavior.
  • the control parameters output by the user and the predicted control parameters of the driver model it is determined whether there is abnormal driving behavior for the vehicle, so that the abnormal driving behavior can be detected without adding new sensors. detection. And because this detection method is directly based on the user's control parameters of the vehicle, it can avoid the limitations of the sensor in detecting abnormal driving behaviors and can also reduce the delay in detecting abnormal driving behaviors.
  • the method further includes: obtaining a historical manipulation parameter sequence, where the historical manipulation parameter sequence includes controlling the traffic within a first historical duration before the first moment. At least one control parameter for tool travel; the driver model is determined based on the historical control parameter sequence.
  • the method further includes: acquiring a second manipulation parameter, a second motion state parameter and second environment information, the second manipulation parameter being used to control the The control parameters of the vehicle driving, the second motion state parameter is used to indicate the movement state of the vehicle at the second moment, and the second environment information is used to indicate the surrounding environment of the vehicle at the second moment; according to The second motion state parameter and the second environment information are used to determine a second expected manipulation parameter based on the driver model; and the determination of whether there is abnormal driving behavior based on the first manipulation parameter and the first expected manipulation parameter includes: Determine whether there is abnormal driving behavior according to a first manipulation parameter sequence and a first expected manipulation parameter sequence, wherein the first manipulation parameter sequence includes the first manipulation parameter and the second manipulation parameter, and the first expected manipulation parameter sequence includes the first expected manipulation parameter and the second expected manipulation parameter.
  • abnormal driving behavior is identified based on the first manipulation parameter sequence and the first expected manipulation parameter sequence, which helps to improve the accuracy of identifying abnormal driving behavior. sex.
  • the method further includes: determining a first confidence level based on the first manipulation parameter sequence and the first expected manipulation parameter sequence, the first confidence level being used to characterize the The degree of similarity between the first manipulation parameter sequence and the first expected manipulation parameter sequence; determining whether there is an abnormal driving behavior based on the first manipulation parameter sequence and the first expected manipulation parameter sequence includes: when the first confidence level is greater than or When equal to the first threshold, it is determined whether there is abnormal driving behavior according to the first operating parameter sequence and the first expected operating parameter sequence.
  • the first confidence level is greater than or equal to the first threshold
  • whether there is an abnormal driving behavior is determined based on the first manipulation parameter sequence and the first expected manipulation parameter sequence, which can avoid environmental information, motion state parameters and driving Anomaly detection caused by driver model operation errors can improve the accuracy of abnormal driving behavior detection.
  • the method further includes: when the first confidence is greater than or equal to a second threshold, based on the first manipulation parameter sequence and the first expected manipulation parameter sequence , the first prediction model is optimized, and the second threshold is smaller than the first threshold.
  • the driver model when the first confidence level is greater than or equal to the second threshold, the driver model is optimized according to the first manipulation parameter sequence, which can make the manipulation parameters output by the driver model more consistent with the user's driving habits. , which can improve the accuracy of abnormal driving behavior detection.
  • the method further includes: obtaining first information, the first information being used to indicate user manipulation preferences; and determining the driver model based on the first information.
  • the method further includes: determining an evaluation function based on the first information; determining whether there is an abnormality based on the first manipulation parameter sequence and the first expected manipulation parameter sequence.
  • the driving behavior includes: determining the value of the evaluation function based on the first control parameter sequence and the first expected control parameter sequence; and determining the presence of abnormal driving behavior when the value of the evaluation function is greater than or equal to a third threshold.
  • the evaluation function is determined based on the first information, so that the evaluation benchmark for abnormal driving behavior can be set according to different user manipulation preferences, thereby reducing the possibility of misjudgment and improving the accuracy of identifying abnormal driving behavior. Spend.
  • the method further includes: determining the working condition type of the vehicle according to the first motion state parameter and/or the first state parameter; Determine the evaluation function based on the working condition type.
  • the corresponding evaluation benchmark for abnormal driving behavior can be set based on various scenarios in the user's actual driving, thereby reducing errors. to improve the accuracy of identifying abnormal driving behavior.
  • the method further includes: when it is determined that the abnormal driving behavior exists, prompting the user that the abnormal driving behavior exists.
  • the method further includes: when it is determined that abnormal driving behavior exists, controlling the vehicle to be in an automatic driving mode.
  • the vehicle is a vehicle
  • the control parameter includes: at least one of a steering wheel angle, an accelerator pedal stroke, and a brake pedal stroke.
  • a device for identifying abnormal driving behavior includes: an acquisition module, which can be used to acquire a first manipulation parameter, a first motion state parameter and a first environment information.
  • the first manipulation parameter is at the first A control parameter for controlling the movement of a vehicle at a moment, the first motion state parameter is used to indicate the motion state of the vehicle at the first moment, and the first environment information is used to indicate the surroundings of the vehicle at the first moment. environment; the processing module may be used to determine a first expected manipulation parameter based on the first motion state parameter and the first environment information based on the driver model; determine whether to determine whether Abnormal driving behavior exists.
  • the method further includes: obtaining a historical manipulation parameter sequence, where the historical manipulation parameter sequence includes controlling the traffic within a first historical duration before the first moment. At least one control parameter for tool travel; the driver model is determined based on the historical control parameter sequence.
  • the acquisition module may also be used to: acquire the second manipulation parameter, the second motion state parameter and the second environment information, where the second manipulation parameter is in the second The control parameters that control the driving of the vehicle at all times, the second motion state parameter is used to indicate the motion state of the vehicle at the second moment, and the second environment information is used to indicate the surroundings of the vehicle at the second moment.
  • the processing module can also be used to: determine the second expected control parameter based on the driver model according to the second motion state parameter and the second environment information; the processing module can be specifically used to: according to the first The operating parameter sequence and the first expected operating parameter sequence determine whether there is abnormal driving behavior, wherein the first operating parameter sequence includes the first operating parameter and the second operating parameter, and the first expected operating parameter sequence includes the first expected handling parameter and the second expected handling parameter.
  • the processing module may also be configured to: determine a first confidence level based on the first manipulation parameter sequence and the first expected manipulation parameter sequence. Used to characterize the degree of similarity between the first manipulation parameter sequence and the first expected manipulation parameter sequence; the processing module is specifically used to: when the first confidence level is greater than or equal to the first threshold, according to the first operating parameter sequence and the first expected manipulation parameter sequence to determine whether there is abnormal driving behavior.
  • the processing module may also be configured to: when the first confidence level is greater than or equal to the second threshold, based on the first manipulation parameter sequence, perform the A prediction model is optimized, and the second threshold is smaller than the first threshold.
  • the acquisition module may also be used to acquire first information, the first information being used to indicate the user's manipulation preferences; the processing module may also be used to obtain the first information according to the third aspect. A piece of information to determine the driver model.
  • the processing module can also be used to: determine the evaluation function according to the first information; specifically, the processing module can be used to: according to the first manipulation parameter sequence and the first expected manipulation parameter sequence to determine the value of the evaluation function; when the value of the evaluation function is greater than or equal to the third threshold, it is determined that abnormal driving behavior exists.
  • the processing module may also be used to: determine the working condition type of the vehicle according to the first motion state parameter and/or the first state parameter; The evaluation function is determined according to the working condition type of the vehicle.
  • the processing module may also be used to: when it is determined that abnormal driving behavior exists, prompt the user that abnormal driving behavior exists.
  • the processing module may also be used to: when it is determined that abnormal driving behavior exists, control the vehicle to be in the automatic driving mode.
  • the vehicle is a vehicle
  • the control parameter includes: at least one of a steering wheel angle, an accelerator pedal stroke, and a brake pedal stroke.
  • a device in a third aspect, includes a processor and a memory, wherein the memory is used to store program instructions, and the processor is used to call the program instructions so that the device executes any possible method in the first aspect.
  • a fourth aspect provides a vehicle, which includes the device in any one of the second aspect or the third aspect.
  • the vehicle may be a smart car or a new energy vehicle.
  • a computer program product includes: computer program code.
  • the computer program code When the computer program code is run on a computer, it causes the computer to execute the method in the first aspect.
  • a computer-readable medium stores program code.
  • the computer program code When the computer program code is run on a computer, it causes the computer to execute the method in the first aspect.
  • a chip in a seventh aspect, includes a processor and a data interface.
  • the processor reads instructions stored in the memory through the data interface, so that the computer executes the method in the first aspect.
  • Figure 1 is a schematic functional block diagram of a vehicle provided by an embodiment of the present application.
  • Figure 2 is a schematic diagram of a driving scenario provided by an embodiment of the present application.
  • Figure 3 is a schematic flow chart of a method for training a driver model provided by an embodiment of the present application.
  • Figure 4 is a schematic flow chart of a method for identifying abnormal driving behavior provided by an embodiment of the present application.
  • Figure 5 is a schematic flow chart of another method for identifying abnormal driving behavior provided by an embodiment of the present application.
  • Figure 6 is a schematic structural diagram of a device for identifying abnormal driving behavior provided by an embodiment of the present application.
  • Figure 7 is a schematic flowchart of a system for identifying abnormal driving behavior provided by an embodiment of the present application.
  • Figure 8 is a schematic flowchart of an optimized driver model provided by an embodiment of the present application.
  • Figure 9 is a schematic flowchart of another method of identifying abnormal driving behavior provided by an embodiment of the present application.
  • Figure 10 is a structural example diagram of a device provided by an embodiment of the present application.
  • vehicles may include one or more different types of vehicles that operate or move on land (eg, highways, roads, railways, etc.), water (eg: waterways, rivers, oceans, etc.) or in space.
  • movable objects e.g., vehicles may include cars, bicycles, motorcycles, trains, subways, airplanes, ships, aircraft, robots, or other types of transportation vehicles or movable objects. The following uses a vehicle as an example to briefly introduce the functions that a vehicle can have.
  • FIG. 1 is a functional block diagram of a vehicle 100 provided by an embodiment of the present application.
  • the vehicle 100 may include a perception system 120 , a display device 130 , and a computing platform 150 , where the perception system 120 may include several types of sensors that sense information about the environment surrounding the vehicle 100 .
  • the sensing system 120 may include a positioning system.
  • the positioning system may be a global positioning system (GPS), a Beidou system or other positioning systems, an inertial measurement unit (IMU), a lidar, a millimeter One or more of wave radar, ultrasonic radar and camera device.
  • GPS global positioning system
  • IMU inertial measurement unit
  • lidar a millimeter One or more of wave radar, ultrasonic radar and camera device.
  • the computing platform 150 may include processors 151 to 15n (n is a positive integer).
  • the processor is a circuit with signal processing capabilities.
  • the processor may be a circuit with instruction reading and execution capabilities.
  • CPU central processing unit
  • microprocessor microprocessor
  • GPU graphics processing unit
  • DSP digital signal processor
  • the processor can realize certain functions through the logical relationship of the hardware circuit. The logical relationship of the hardware circuit is fixed or can be reconstructed.
  • the processor is an application-specific integrated circuit (application-specific integrated circuit).
  • Hardware circuits implemented by ASIC or programmable logic device (PLD), such as field programmable gate array (FPGA).
  • PLD programmable logic device
  • FPGA field programmable gate array
  • the process of the processor loading the configuration file and realizing the hardware circuit configuration can be understood as the process of the processor loading instructions to realize the functions of some or all of the above units.
  • it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a neural network processing unit (NPU), tensor processing unit (TPU), deep learning processing Unit (deep learning processing unit, DPU), etc.
  • the computing platform 150 may also include a memory, and the memory may be used to store instructions.
  • processors 151 to 15n may call instructions in the memory and execute the instructions to implement corresponding functions.
  • the memory may also be used to store data such as road maps, route information, vehicle location, direction, speed and other such vehicle data, as well as other information. This information may be used by vehicle 100 and computing platform 150 during operation of vehicle 100 in autonomous, semi-autonomous and/or manual modes.
  • the vehicle 100 may include an advanced driving assist system (ADAS), which utilizes a variety of sensors on the vehicle (including but not limited to: lidar, millimeter wave radar, camera, ultrasonic sensor, global positioning system, inertial Measurement unit) acquires information from around the vehicle and analyzes and processes the acquired information to achieve functions such as obstacle perception, target recognition, vehicle positioning, path planning, driver monitoring/reminder, etc., thereby improving the safety of vehicle driving. Level of automation and comfort.
  • ADAS advanced driving assist system
  • FIG. 2 is a schematic diagram of a driving scenario provided by an embodiment of the present application.
  • the driver can control the vehicle by sensing environmental information.
  • the vehicle can include a body system, a power transmission system, a chassis system, etc., thereby controlling vehicle acceleration, braking, steering, etc.
  • the vehicle can be
  • the vehicle 100 shown in Figure 1 can obtain information about its environment through its sensing system. For example, it can obtain information about roads, road fences, intersections and other traffic facilities in its environment; it can obtain information about the road surface used to divide lanes.
  • Information on road conditions such as lane lines, road potholes, and road slopes; information on traffic participants such as the location, size, and speed of surrounding pedestrians and vehicles can be obtained, and information on cloudy, sunny, rainy, and snowy days can be obtained Information about weather conditions, etc. will not be repeated here for the sake of brevity.
  • Figure 3 is a schematic flow chart of a method for training a driver model provided by an embodiment of the present application.
  • the method 200 may include;
  • the first manipulation parameter may be a manipulation parameter for controlling the movement of the vehicle. That is to say, the first manipulation parameter may be a manipulation parameter output by the user for controlling the movement of the vehicle.
  • the first control parameter may be a user-controlled control parameter of the vehicle when the vehicle is in manual driving mode, such as the user's control of the steering wheel, accelerator pedal, brake pedal and other controllable tools.
  • the corresponding information outputted is the steering wheel angle, accelerator pedal stroke, brake pedal stroke and other information.
  • the first manipulation parameter may also include, for example, clutch pedal stroke, etc. It should be understood that the above method regarding the first manipulation parameter is only an example, and the embodiments of the present application do not limit this.
  • the motion state information can be used to characterize the motion state of the vehicle.
  • the motion state information can be used to characterize the motion state of the vehicle.
  • the motion status information may include the vehicle's speed, acceleration, angular velocity, angular acceleration, etc.
  • the motion status information of the vehicle may be determined based on data collected by sensors of the sensing system.
  • the vehicle's sensing system may include an IMU, and based on the collection results of the IMU, the acceleration and/or angular acceleration of the vehicle may be determined; for another example, the vehicle may include a vehicle speed sensor, based on The data collected by the vehicle speed sensor can determine the speed of the vehicle; for another example, the vehicle can also include a wheel speed sensor, and the speed of the vehicle can be determined based on the data collected by the wheel speed sensor. It should be understood that the above methods of obtaining the motion status information of the vehicle are only examples for ease of explanation, and are not limited in the embodiments of the present application.
  • the environment information can be used to describe the environment information where the vehicle is located.
  • the information of other traffic elements such as vehicles and pedestrians around the vehicle can be determined based on the data collected by the vehicle's perception system. Based on the information of the traffic elements, it can be determined that the vehicle is driving. Obstacles that may exist in the vehicle can be obtained. For example, the area where other surrounding vehicles are located, the heading angle and distance of the surrounding vehicles relative to the vehicle can be obtained, so that the vehicle's driving path can be planned; the vehicle's sensors and/or Internet of Vehicles data can be used to plan the vehicle's driving path. , determine the information of the road where the vehicle is located.
  • the road information can be the slope, inclination angle, width of the road and other information. It should be understood that the environmental constraints encountered by the vehicle during driving can be determined based on the environmental information. It should be understood that the above description of environmental information is only an example for ease of explanation, and the embodiments of the present application do not limit this.
  • steps S210 to S230 can be performed at the same time, or step S210 can be performed first, S220 can be performed first, or S230 can be performed first. This is not limited in the embodiments of the present application.
  • S240 Carry out model training based on the first control parameter, motion state information and environmental information to obtain a driver model.
  • the driver model can be used to predict and control the control parameters of the vehicle.
  • the first manipulation parameter has a corresponding relationship with the motion state parameter and environmental information.
  • the first control parameter may be a control parameter that controls the movement of the vehicle at any time
  • the first motion state information may indicate the motion state of the vehicle at that time
  • the environmental information may indicate the state of the vehicle at that time.
  • the surrounding environment where the user is located that is to say, at that moment, the user's control of the vehicle can be reflected in the first control parameter according to the motion state of the vehicle and the surrounding environment. Therefore, the user can control the vehicle according to the first control parameter.
  • the corresponding motion state parameters and environmental information are used for model training, so that a driver model for predicting and controlling the control parameters of the vehicle can be obtained.
  • the driver model obtained based on this training can output control parameters to simulate autonomous driving based on the vehicle's perception of the real-time environment and its own motion state.
  • the acquired motion state parameters and environmental information may not, in a strict sense, indicate the motion of the vehicle at the moment when the first control parameter is located. state and the surrounding environment.
  • the motion state can be considered
  • the time corresponding to the parameter, the time corresponding to the environment information, and the time corresponding to the first control parameter are the same time. For example, it is intended to obtain the first control parameters, motion state parameters and environmental information at the 10th second (second, s).
  • the first control parameter obtained may be at the 10.05s
  • the obtained motion state parameters may be the motion state of the vehicle at the 9.91s
  • the corresponding environmental information obtained may be the surrounding environment of the vehicle at the 10.03s. Since the time difference between the corresponding moments is less than or equal to the preset time difference threshold (such as 0.1s), it can be considered that the obtained motion state parameters and environmental information indicate the motion state of the vehicle at the moment corresponding to the first control parameter.
  • the time corresponding to the first control parameter, the motion state parameter and the surrounding environment is the same time (for example, the 10th s), that is, it can be considered that the obtained motion state parameter indicates that the vehicle is in the 10th s.
  • the environmental information indicates the surrounding environment of the vehicle at the 10th second.
  • the first control parameter can be considered to be the control parameter that controls the movement of the vehicle at the 10th second.
  • the driver model may be obtained by performing model training based on at least one set of first manipulation parameters and corresponding at least one set of motion state information and environmental information. For example, determining the driver model based on multiple sets of first control parameters, and corresponding motion state information and environmental information can be based on optimization algorithms, such as tabu algorithm, genetic algorithm, particle swarm optimization algorithm, ant colony algorithm, etc. , so as to obtain a driver model that matches the user's actual maneuvering behavior. For the sake of brevity, they will not be described here.
  • the manipulation of the vehicle may be different in the same driving scenario.
  • driving scenarios such as overtaking, lane changing, etc.
  • driving scenarios such as overtaking, lane changing, etc.
  • there may be large differences in the operation of the steering wheel, accelerator pedal, brake pedal, gear, etc. so in the same scene, the driving of the two
  • the vehicles can have different motion states (such as different speeds, accelerations, etc.), which can be reflected in the fact that the vehicles driven by the two can have different driving trajectories, etc.
  • driver models can be obtained through model training, so that different driver models can be obtained according to different driving preferences.
  • the driver model predicts the driving behavior of users with different driving preferences to improve the accuracy of prediction. For example, taking a vehicle as a means of transportation, for example, for users with conservative driving preferences, a preview driver model can be used, while for users with aggressive driving preferences, a driver model based on neural network training can be used.
  • driver models can be used for users with different driving styles; for example, for users with conservative driving preferences, a driver model based on neural network training can be used, while for users with aggressive driving preferences, Driver models based on neural network training can also be used.
  • the above two driver models can have different characteristic parameters. That is to say, for users with different driving styles, the same type of driver models with different characteristic parameters can be used.
  • Driver model can be used for users with different driving styles; for users with different driving styles; the same type of driver models with different characteristic parameters can be used.
  • the driver model can also be optimized and adjusted based on the first steering parameter, motion state information and environmental information.
  • the optimization of the driver model and adjustments can include multiple implementations.
  • this process may include multiple iterative processes.
  • An iterative process may include the following steps:
  • S310 Input the motion state information and environment information to the driver model, process the motion state information and environment information through the driver model, and output the prediction results, that is, output the predicted control parameters of the vehicle.
  • the first loss value can be calculated through the loss function.
  • the first loss value can represent the deviation between the prediction result and the first control parameter.
  • the prediction result and the first control parameter The greater the deviation between , the greater the first loss value.
  • the above shows an iterative process in optimizing the driver model. After an iteration is performed, it can be detected whether the optimization termination condition has been met. When the termination condition is not met, the next iteration process is performed; when the optimization termination condition is met, The driver model used in this iteration process can be output as a driver model that is ultimately used to predict the driver's control parameters.
  • the optimization termination condition can be that the number of iterations reaches the target number or the loss function meets the preset conditions, or it can also be that the ability does not improve within a certain number of iterations.
  • the target number can be a preset number of iterations to determine the end time of the optimization process of the driver model to avoid wasting training resources;
  • the termination condition can be that during the iteration process, the loss function value within a period of time remains unchanged or does not decrease. For example, if the absolute value of the difference between two or more consecutive loss functions is less than the target value, it means that the adjustment of the driver model has achieved the desired effect.
  • the driver model can be determined by the cloud server.
  • the cloud server can also implement optimization of the driver model.
  • the vehicle can obtain the driver model determined/optimized by the cloud server; it can also be determined by the cloud server. After the driver model is determined, the vehicle obtains the driver model, and can subsequently optimize the driver model, or the vehicle can determine the driver model and optimize the driver model. This application The embodiment does not limit this.
  • the user's control parameters of the vehicle can be predicted based on the driver model, so that abnormal driving behavior can be detected based on the predicted control parameters.
  • Figure 4 is a schematic flow chart of a method for identifying abnormal driving behavior provided by an embodiment of the present application.
  • the method 400 includes the following steps:
  • the first moment can be any moment
  • the control parameters that control the movement of the vehicle at that moment can be obtained, that is, the first control parameter.
  • the first control parameter can represent the user's control of the traffic at the first moment.
  • the first motion state parameter may be used to indicate the motion state of the vehicle at the first moment
  • the first environment information may be used to indicate the surrounding environment where the vehicle is located at the first moment.
  • control parameters that control the movement of the vehicle may be referred to as active control parameters.
  • the user's abnormal driving behavior can be continuously detected.
  • the active control parameters within a preset period of time, as well as the corresponding motion status information and environmental information within the period of time can be obtained to determine whether there is abnormal driving behavior within the period of time; for another example, in order to save the occupied resources, the system can be based on a certain
  • the sampling frequency is used to obtain multiple sets of active control parameters and corresponding motion state information and environmental information within the preset time period.
  • the multiple sets of control parameters can form a sequence in time order.
  • the embodiment of the present application distinguishes active control parameters at multiple moments.
  • the active control parameters at the first moment can be defined as the first control parameter, and accordingly, will be used to indicate traffic
  • the movement state parameter of the movement state of the tool at this moment is defined as the first movement state parameter.
  • the environmental information used to indicate the surrounding environment of the vehicle at this moment is defined as the first environment information.
  • the active state at the second moment can be defined as the first environment information.
  • the manipulation parameters are defined as second manipulation parameters.
  • the motion state parameters and environmental information at that moment can be defined as second motion state parameters and second environment information.
  • the operating parameters used to control the driving of the vehicle within the first historical period before the first moment can be called historical operating parameters, and the sequence of at least one historical operating parameter within the first historical period can be called Historical control parameter sequence.
  • the historical control parameters can be the control parameters that control the driving of the vehicle during this trip, or the control parameters that control the driving of the vehicle during the user's last drive.
  • the first historical duration can be any length, such as 5 minutes, 10 minutes, etc. For the sake of brevity, no further details will be given here.
  • active manipulation parameters at multiple moments can be obtained, as well as motion state parameters and environmental information corresponding to the multiple moments.
  • a first manipulation parameter sequence, a motion state parameter sequence and an environment information sequence can be obtained.
  • the first manipulation parameter sequence can include active manipulation parameters at multiple times, for example, it can include a first manipulation parameter and a second manipulation parameter.
  • the parameters may also include active control parameters at other times.
  • the motion state parameter sequence may include corresponding motion state parameters at the multiple times.
  • the environment information sequence may include corresponding environmental information at the multiple times.
  • multiple active control parameters within the time window can be obtained based on the time window.
  • the multiple active control parameters can constitute a first control parameter sequence in a certain order (such as time sequence, etc.).
  • the time can be determined.
  • the duration of the time window may be a fixed duration, or may be a duration determined based on the driving scene, etc. This embodiment of the present application does not limit this.
  • first information can be obtained, and the first information can be used to indicate the user's manipulation preferences.
  • the first information can be determined according to the user identification.
  • the user identification can be obtained, for example, by obtaining the user's voiceprint information, facial features, obtaining the account information used by the user, etc., to determine the user identification.
  • the server can be queried on the server or the local database.
  • the user manipulation preference related to the user identification is determined, so that the first information can be determined based on the user identification. It should be understood that the method of obtaining the user identification can refer to the existing technology, and for the sake of simplicity, it will not be described again here.
  • the first information may be determined based on historical manipulation data. For example, taking a vehicle as a means of transportation, the user's driving style can be determined based on the user's manipulation of the vehicle within a period of time. For example, the expected driving paths and maneuvers of users with different driving styles in different scenarios can be preset. Data, etc., can be matched according to the user's vehicle manipulation data within a certain period of time (such as 5 minutes) with the preset expected manipulation data of different driving styles, so that the first information can be determined; for another example, the same user There may also be differences in driving preferences in different scenarios. After the working conditions of the vehicle change, it can be determined whether the first information has changed based on the user's manipulation of the vehicle within a certain period of time. When the first information changes, When, accordingly, the driver model can be determined, thereby ensuring the accuracy of the abnormal behavior detection method.
  • the first information can be determined according to user information. For example, taking a vehicle as a means of transportation, when the user drives the vehicle for the first time, he or she can set his or her driving preferences through terminals such as mobile phones or tablets, or cockpit screens such as the vehicle central control screen; for another example, when the user is driving the vehicle , user information such as gender, age, driving years, etc. can be input through terminals such as mobile phones, tablets, and cockpit screens such as vehicle central control screens.
  • the matching user information can be determined based on the user characteristics table stored locally or on the server.
  • the user characteristics table can be used to determine the mapping relationship between the user information and the first information. For example, the user characteristics can be determined based on the analysis of the manipulation preferences of different users of different ages, genders, driving years, education levels, etc. table, which will not be described here for the sake of brevity.
  • the first information can be in any data format, such as numbers, letters, character strings, etc.
  • labels may be used to represent the manipulation preference.
  • labels such as “aggressive”, “conservative” and “neutral” can be used to represent the first information, where the label "aggressive” can indicate that the user's manipulation method is more aggressive.
  • the vehicle for example, when driving a vehicle At a traffic light intersection, when the green light time is short, users with aggressive maneuvering methods may increase their vehicle speed to pass the intersection.
  • users with conservative maneuvering methods may reduce their vehicle speed and wait for the next traffic light.
  • numbers, letters, and characters can be used to represent different driving styles.
  • the numbers “1", “2”, and “3” can be used to represent users with different driving styles.
  • the number “1” represents users with aggressive driving preferences
  • the number "5" represents users with more conservative driving styles. user. It should be understood that the above description of driving styles is only an example for ease of explanation, and different manipulation preferences can also be distinguished in other ways, which is not limited in the embodiments of the present application.
  • the operations/behaviors such as the acquisition, storage, utilization, and processing of user information involved in this application are all used in compliance with local laws and regulations.
  • the operations such as obtaining, storing, utilizing, and processing user information involved in this patent represent operations with the consent of the individual.
  • the driver model can be determined based on the first information.
  • a matching driver model can be determined based on the first information. For example, a corresponding matching driver model is determined based on the first information, such as in a query manner. For the sake of brevity, details will not be described here.
  • the first information can be obtained, and the driver model can be determined based on the first information. This can improve the accuracy of abnormal driving behavior detection.
  • the first expected manipulation parameter may be determined.
  • the user's manipulation parameters for the vehicle can be predicted under the motion state and the surrounding environment.
  • Parameters may refer to steering parameters determined based on the driver model.
  • the expected maneuvering parameters at that moment can be determined.
  • the traffic can be planned based on the motion state parameters and environmental information.
  • the target speed and/or target path of the tool in this scenario so that based on the target speed and/or target path, the driver model can output the expected control parameters to predict the user's behavior when the vehicle is in this motion state and environment.
  • the maneuvering of the vehicle that is, the first expected maneuvering parameter. It should be understood that the above method for determining the first expected manipulation parameter is only an example for ease of explanation, and is not limited in this embodiment of the present application.
  • multiple corresponding expected manipulation parameters can be determined based on the driver model.
  • the first expected manipulation parameter sequence can be determined, and the first expected manipulation parameter sequence can be Includes multiple expected manipulation parameters.
  • the driver model can be determined based on historical manipulation parameter sequences. For example, while the vehicle is driving, the active control parameters can be obtained and recorded in real time.
  • the matching driver model can be determined based on the user's control of the vehicle within a certain period of time.
  • the driver model can also be selected based on the active control parameters within that time.
  • the operator model is optimized and adjusted to ensure the accuracy of the abnormal behavior detection method.
  • S450 Determine whether there is abnormal driving behavior based on the first manipulation parameter and the first expected manipulation parameter.
  • the vehicle is a vehicle and the control parameters are the accelerator pedal stroke, the brake pedal stroke and the steering wheel angle
  • the accelerator pedal stroke is 15 centimeters (centimeter, cm)
  • the brake pedal stroke is The stroke is 0 and the steering wheel angle is 0, that is, the user does not control the vehicle to decelerate in this scenario.
  • the accelerator pedal stroke is 0 cm
  • the brake pedal stroke is 15 cm
  • the steering wheel The turning angle is 0, that is, the driver model predicts that the user needs to control the vehicle to decelerate in this scenario.
  • the first control parameter and the first expected control parameter can be determined by looking up the table.
  • the difference between the two is greater than or equal to the threshold, it can be determined that abnormal driving behavior exists.
  • the set threshold can be preset in advance, or calculated based on the working conditions of the vehicle. This application applies This is not limited; for another example, the value of the evaluation function can be determined based on the multiple active control parameters and the expected control parameters based on the evaluation function, thereby determining whether there is abnormal driving behavior.
  • the first manipulation parameter may include the steering wheel angle, accelerator pedal stroke, brake pedal stroke, etc. output by the user.
  • the first expected manipulation parameter may also include the driver model.
  • the first manipulation parameter is represented by ( ⁇ , ⁇ , F b ), where ⁇ can be used to represent the steering wheel angle output by the user, and ⁇ can be used to represent the accelerator pedal stroke output by the user.
  • F b can be used to represent the brake pedal stroke output by the user.
  • the accelerator pedal stroke and brake pedal stroke can be the actual pedal stroke, or they can be the normalized pedal stroke (such as the total pedal stroke for the user The output pedal stroke is normalized, etc.).
  • the first expected control parameter is represented by ( ⁇ m , ⁇ m , F bm ), where ⁇ m can be used to represent the steering wheel angle determined by the driver model, ⁇ m can be used to represent the accelerator pedal stroke determined by the driver model, and F bm can be used to represent the brake pedal stroke determined by the driver model.
  • the first control parameter and the first expected control parameter can also be other control parameters such as the vehicle gear position. For the sake of simplicity, examples will not be explained one by one here. It should be understood that the above for the first control parameter and the first expected control parameter The representation method is only an example for convenience of explanation, and the embodiment of the present application does not limit this.
  • a first confidence level may be determined, and the first confidence level may be used to represent the degree of similarity between the first manipulation parameter and the first expected manipulation parameter.
  • the length of the residual vector can be mapped to the interval [0,1] to represent the degree of similarity between the first control parameter and the first expected control parameter, that is to say , the first confidence level can be a function of the length of the residual vector. For example, let ⁇ represent the first confidence level, The greater the first confidence ⁇ , the higher the degree of similarity between the first manipulation parameter and the first expected manipulation parameter.
  • the above methods of determining the first confidence level are only examples for ease of explanation, and other methods can be used to determine the first confidence level.
  • multiple control parameters can be determined based on the impact of steering wheel angle, accelerator pedal stroke and other control parameters on driving safety.
  • Different weights are set, and the first confidence level is determined based on the weights. It should be understood that the embodiment of the present application does not limit the method of determining the first confidence level.
  • the first confidence level when the first confidence level is greater than or equal to the first threshold, it may be determined according to the first manipulation parameter and the first expected manipulation parameter whether there is an abnormal driving behavior.
  • the first confidence level when the first confidence level is greater than or equal to the first threshold, the consistency between the first control parameter and the first expected control parameter is good, and the driver model can be considered to be in a normal operating state. Therefore, according to the first The first control parameter and the first expected control parameter determine whether there is abnormal driving behavior, which can reduce misjudgments of abnormal driving behavior.
  • whether there is abnormal driving behavior may be determined based on a plurality of active manipulation parameters and a plurality of expected manipulation parameters within a first period of time.
  • the first duration may be the duration of the sampling time window of the active manipulation parameters, or the duration of the time window used for detecting abnormal driving behavior.
  • the plurality of first manipulation parameters and the plurality of first expected manipulation parameters The numbers correspond one to one.
  • abnormal driving behavior within the first time period can be detected based on multiple active control parameters and corresponding expected control parameters within the first time period. Since the user's manipulation of the vehicle is usually consistent, the detection and identification of abnormal driving behavior based on the multiple manipulation parameters output by the user within the first period of time helps to improve the accuracy of detecting and identifying abnormal driving behavior.
  • the user's abnormal driving behavior can be detected based on a time window.
  • the duration of the time window can be the first duration.
  • the user can output multiple sets of Active control parameters, correspondingly, the driver model can also output multiple sets of expected control parameters.
  • the plurality of sets of active control parameters may constitute a first control parameter sequence in a preset sorting manner.
  • the plurality of corresponding sets of predicted control parameters may constitute a first expected control parameter sequence.
  • the first control parameter sequence The first expected control parameter sequence may have the same sorting method, that is to say, the first control parameter sequence may include active control parameters of N times, and the first expected control parameter sequence may include N expected control parameters of time.
  • N is a positive integer greater than or equal to 2.
  • ( ⁇ i , ⁇ i , F bi ) can be used to represent the i-th group of active control parameters in the first control parameter sequence (for example, it can be the active control at the first moment parameters)
  • ( ⁇ mi , ⁇ mi , F bmi ) can be used to represent the i-th group of expected control parameters in the first expected control parameter sequence (for example, it can be the first moment predicted based on the driver model.
  • the sequence of multiple residual vectors composed according to the order of multiple sets of active control parameters can be referred to as the residual. Sequence, where i is a positive integer less than or equal to N.
  • the first confidence level may also be used to represent the degree of similarity between the multiple sets of active control parameters and the multiple sets of expected control parameters.
  • the first period of time may include n sets of active control parameters ( ⁇ i , ⁇ i , F bi ) and expected control parameters ( ⁇ mi , ⁇ mi , F bmi ), correspondingly , the residual vector can be determined Among them, n is a positive integer greater than or equal to 2, and i is a positive integer less than or equal to n.
  • the degree of similarity between the multiple sets of active control parameters and the multiple sets of expected control parameters can be determined based on the Euclidean distance.
  • the first confidence level ⁇ may be a function of the Euclidean distance d.
  • the Euclidean distance can be mapped to the interval [0,1] to characterize the degree of similarity between the multiple sets of active control parameters and the multiple sets of expected control parameters, that is, to determine the first confidence level, For the sake of brevity, no further details will be given here.
  • whether there is abnormal driving behavior may be determined based on the plurality of sets of first manipulation parameters and the plurality of sets of second parameters within the first time period.
  • the first confidence level is greater than or equal to the first threshold
  • whether there is an abnormal driving behavior is determined based on the first manipulation parameter sequence and the first expected manipulation parameter sequence, which can avoid environmental information, motion state parameters and driving Anomaly detection caused by driver model operation errors can improve the accuracy of abnormal driving behavior detection.
  • the driver model when the first confidence level is greater than or equal to a second threshold, the driver model may be optimized according to the first manipulation parameter sequence, and the second threshold may be less than or equal to the first threshold. For example, when it is determined that there is no abnormal driving behavior and the first confidence level is greater than or equal to the second threshold, the driver model can be optimized according to the first manipulation parameter sequence, so that the expected manipulation parameters determined by the driver model can be It is more consistent with the user's driving behavior, thereby improving the accuracy of abnormal driving behavior detection.
  • determining whether there is abnormal driving behavior based on the first manipulation parameter sequence and the first expected manipulation parameter sequence may be based on the residual sequence and the evaluation function to determine whether there is abnormal driving behavior.
  • the active control parameters and the expected control parameters can be functions that change with time, such as the active control parameters ( ⁇ (t), ⁇ (t), F b (t)), the expected control parameters
  • the manipulation parameters ( ⁇ m (t), ⁇ m (t), F bm (t)), and the start and end times of the time window are represented by t 1 and t 2 respectively.
  • the residual vector can also be a function that changes with time, It can be determined whether there is abnormal driving behavior based on the evaluation function.
  • the evaluation function can be Among them, k 1 , k 2 , and k 3 can be the weight coefficients of each control parameter, and J can be the value of the evaluation function, or called the first evaluation value.
  • the working condition of the vehicle can be determined based on at least one of motion state information and environmental information.
  • the vehicle's motion state such as the vehicle's speed, acceleration, angular velocity, etc.
  • environmental information such as the road gradient, the expected vehicle speed on the road, etc.
  • the first control parameter such as One or more of the steering wheel angle, accelerator pedal stroke, brake pedal stroke, etc.
  • the evaluation function can be in different functional forms.
  • the evaluation function can be a linear function, a quadratic function, a power function, an exponential function, etc.
  • the above method of determining the evaluation function according to the working conditions of the vehicle is only an example, and the embodiments of the present application do not limit this.
  • the evaluation function may be determined based on the first information.
  • the driver model can be determined based on the user's manipulation preference.
  • the evaluation function can be determined based on the first information.
  • the evaluation function can be determined by combining the working conditions and the first information. For example, after determining the operating conditions of the vehicle and the first information, the corresponding evaluation function can be determined based on a two-dimensional table lookup method, etc., which will not be described again here for the sake of simplicity.
  • whether there is abnormal driving behavior may be determined based on an evaluation function based on a plurality of sets of first manipulation parameters and second manipulation parameters acquired within a first period of time. For example, taking the vehicle as a means of transportation, the duration of the sampling time window as the first duration, and its start and end times as t 1 and t 2 respectively, the evaluation function can be Whether there is abnormal driving behavior can be determined based on the evaluation function based on the residuals between the multiple sets of first control parameters and the second control parameters within the time window, for example, when the calculation result of the evaluation function is greater than or equal to the third threshold.
  • the second time period can include one or more sampling time windows . It should be understood that the above methods for determining abnormal driving behavior are only examples, and the embodiments of the present application do not limit this.
  • the control parameters output by the user and the control parameters of the driver model it is determined whether there is abnormal driving behavior for the vehicle, so that the detection of abnormal driving behavior can be realized without adding new sensors. .
  • this detection method is directly based on the user's control parameters of the vehicle, it can avoid the limitations of the sensor in detecting abnormal driving behaviors and can also reduce the delay in detecting abnormal driving behaviors.
  • differentiated configuration of the driver model can be achieved, and the accuracy of the abnormal driving behavior detection method can be improved.
  • the expected control parameters can be regarded as the control parameters output when the user normally controls the vehicle, and the expected control parameters are obtained based on the driver model, which can avoid storage space, communication quality, database acquisition, etc.
  • the driver model is determined through historical manipulation sequences, which can improve the accuracy of identifying abnormal driving behaviors.
  • the driver model is determined based on the user's manipulation preferences, which can also achieve differentiation of driver models. Settings can also improve the accuracy of identifying abnormal driving behavior.
  • Figure 5 is a schematic flowchart of a method for identifying abnormal driving behavior provided by an embodiment of the present application.
  • the method 500 includes some or all of the following steps:
  • the active manipulation parameters for controlling the driving of the vehicle can be obtained.
  • the first control parameter may include steering wheel angle, accelerator pedal stroke, brake pedal stroke, etc.
  • the vehicle may adjust its speed, acceleration, angular velocity and other motion states according to the first control parameter.
  • a plurality of active manipulation parameters output by the user within a first period of time may be obtained.
  • a plurality of active control parameters within the first time period may constitute a first control parameter sequence.
  • the number of active manipulation parameters included in the first manipulation parameter sequence may be greater than or equal to the preset threshold.
  • the preset threshold can also be other values (such as 3, 10, etc.), which is not limited in the embodiments of this application.
  • the description of the vehicle's motion status information and environmental information may refer to steps S220 and S230. For the sake of brevity, they will not be described again here.
  • corresponding multiple motion state parameters and environmental information can be acquired.
  • the active control parameters, vehicle motion state information and environmental information can be acquired at the same time, or the active control parameters can be acquired first, the motion state information can be acquired first, or the environment information can be acquired first.
  • This application implements This example does not limit this.
  • S516 Obtain the first information and determine the driver model based on the first information.
  • the first information may be determined based on historical manipulation parameters.
  • the historical manipulation data may be active manipulation parameters previously stored in the database, such as the driving data retained by the user when he drove the vehicle last time; the historical manipulation parameters may also be generated during this driving process.
  • the user has been driving the vehicle for 15 minutes, and the active control parameters within 10 minutes of this drive can be obtained.
  • the user's control preferences can be determined through comparison, matching, etc. , thereby determining the first information.
  • the user's control preferences can be determined through comparison, matching, etc. , thereby determining the first information.
  • a driver model may be determined based on the first information.
  • S520 Determine the first expected manipulation parameter sequence based on the vehicle's motion status information and environmental information and based on the driver model.
  • step S430 the description about determining the expected manipulation parameters may involve step S430, which will not be described again here for the sake of brevity.
  • multiple expected control parameters can be determined based on multiple sets of environmental information and vehicle motion state information based on the driver model, and the multiple sets of expected control parameters can be constitute a first expected manipulation parameter sequence.
  • first expected manipulation parameter sequence For the sake of brevity, no further details will be given here.
  • S525 Determine the residual sequence according to the first control parameter sequence and the first expected control parameter sequence.
  • multiple residual vectors may be determined, and the residual vectors may be used to represent the relationship between the set of first control parameters and the corresponding second control parameters. correspondingly, the multiple residual vectors can be formed into a residual sequence based on the sorting of the first manipulation parameter sequence, which will not be described again here for the sake of brevity.
  • the first confidence level may be determined based on the residual sequence.
  • the residual sequences of the multiple sets of active control parameters ( ⁇ i , ⁇ i , F bi ) and expected control parameters ( ⁇ mi , ⁇ mi , F bmi ) can be expressed as Represented, for example, according to the Euclidean distance between the residual vector and the zero point Characterizes the degree of similarity between the active control parameters and the expected control parameters.
  • step S535 Determine whether the first confidence level is greater than or equal to the first threshold. If the first confidence level is greater than the first threshold, jump to step S545.
  • the description of the first threshold may refer to step S450, and for the sake of brevity, it will not be described again here.
  • S540 determine the working condition of the vehicle based on the first operating parameter, environmental information and motion state information of the vehicle. For example, based on the speed and acceleration of the vehicle, it can be determined whether the vehicle is in an accelerating condition; based on the environmental information, the distribution of obstacles around the vehicle can be determined, and combined with the vehicle's motion status information, it can be determined whether the vehicle is changing lanes. Obstacle avoidance conditions, etc. It should be understood that the description of the working conditions and working condition types of the vehicle can refer to related technologies. For the sake of simplicity, examples will not be given one by one here.
  • the evaluation function can be determined according to the working conditions.
  • the form of the evaluation function can be determined by looking up a table according to the working conditions, or the weights of multiple control parameters in the evaluation function can be determined according to the working conditions.
  • examples will not be given one by one here.
  • the evaluation function may be determined based on the first information, or the evaluation function may be determined based on the first information combined with the working conditions. For the sake of simplicity, details will not be described here.
  • step S450 the description of the evaluation function may refer to step S450, which will not be described again here for the sake of brevity.
  • step S550 Determine whether there is abnormal driving behavior. When abnormal driving behavior exists, step S555 can be jumped.
  • the first evaluation value can be determined through an evaluation function. For example, taking the duration of the sampling time window as the first duration, its start and end times are t 1 and t 2 respectively, and the residual sequence is
  • the evaluation function is For example, at least one of the weight coefficients k 1 , k 2 , and k 3 can be determined according to the working conditions and/or the first information, so that the evaluation function can be determined, and the first evaluation function can be determined according to the residual sequence.
  • the second time period may include multiple sampling time windows of the first time period (for example, (including 5 sampling time windows), the corresponding first evaluation value can be determined according to the residual sequence of each first duration sampling time window, when the number of the plurality of first evaluation values greater than or equal to the third threshold is not When it is lower than the preset number (for example, the preset number is 3), for example, if 3 of the first evaluation values in the 5 sampling time windows are greater than or equal to the third threshold, it can be determined that there is abnormal driving behavior in the second period of time; For another example, if there are multiple sampling time windows in which the first evaluation value is greater than or equal to the third threshold within the second period of time and the number is greater than a preset number (for example, the preset number is 2), it can be determined that there is abnormal driving behavior. . It should be understood that the above method of determining whether abnormal driving behavior exists is only an example
  • S555 prompts the user that there is abnormal driving behavior.
  • the user may be alerted.
  • text, images, etc. can be displayed on the screen in the vehicle to remind the user.
  • the words "Please pay attention to driving safety” are displayed on the central control screen of the vehicle.
  • the head-up display displays the words "Please drive safely” on the front windshield of the vehicle, and can also remind the user through voice, etc.
  • the first function when it is determined that abnormal driving behavior exists, it can be used to trigger the first function.
  • the first function may be an automatic driving function, an advanced assisted driving function, etc.
  • the user's manipulation of the vehicle may be recognized as abnormal driving behavior due to limitations such as blind spots in the field of vision, and Since the vehicle's perception system can better realize the perception of the surrounding environment, when it is determined that there is abnormal driving behavior, it can directly trigger the automatic driving function or advanced assisted driving function, or it can also prompt through voice reminder or central control screen.
  • the automatic parking function can be triggered to control the vehicle to automatically park on a safe road section; for another example, when the chip and the abnormal driving behavior identification device determine that abnormal driving behavior exists, the automatic parking function can be triggered.
  • the chip or abnormal behavior recognition device may directly trigger the first function, or may send a message to other devices to instruct the other devices to trigger the first function. For the sake of simplicity, details will not be described here.
  • step S560 determine whether the first confidence level is greater than or equal to the second threshold. If the first confidence level is greater than or equal to the second threshold, step S565 can be skipped.
  • the driver model may be optimized according to the first manipulation parameter sequence.
  • the description of the second threshold may refer to step S440, which will not be described again for the sake of brevity.
  • step S560 and step S535 can be performed at the same time, or step S535 can be performed first, or step S560 can be performed first, which is not limited in the embodiment of the present application.
  • the driver model can be optimized based on the residual sequence based on an optimization algorithm.
  • the driver model can be optimized based on the tabu algorithm, genetic algorithm, particle swarm optimization algorithm, and ant colony algorithm. Wait, no more examples here.
  • the positions of particles in the N-dimensional space can respectively represent the control parameters that need to be optimized.
  • V id ⁇ V id +C 1 radm (0,1)(P id -X id )+C 1 radm(0,1)(P gd -X id ),
  • X id X id +V id updates the speed and position of each particle, where C 1 , C 2 It can be an acceleration factor, radm(0,1) can be a random number between the interval (0,1), V id can be the velocity vector of the i-th particle on the d-th dimension, and X id can represent the i-th particle.
  • the position vector on the d-th dimension, ⁇ is the inertia factor, its value is non-negative
  • P id is the position vector of the d-dimensional variable of the optimal solution of the i-th particle in the iterative process
  • P gd is the optimal solution among all particles.
  • the position vector of the d-th dimension variable of the optimal solution, d is a positive integer.
  • the driver model may be optimized through a loss function according to the first manipulation parameter sequence and the first expected manipulation parameter.
  • the residual sequence can be determined based on the first control parameter sequence and the control parameters predicted by the driver model optimized by the particle swarm algorithm, and the first loss value can be determined through a loss function.
  • the loss function can be Wherein, cost can be the first loss value, and f can be the functional relational expression of mapping the residual to the first loss value, that is, the loss function, thereby adjusting the driver model according to the first loss value. It should be understood that the description of the loss function can refer to related technologies, and will not be repeated here for the sake of simplicity.
  • the optimization of the driver model can be implemented by the cloud server, or the optimization of the driver model can be implemented by the vehicle.
  • the first manipulation parameter sequence and the corresponding plurality of environmental information and motion state parameters can be sent to the cloud server, so that the cloud server can implement the control of the driver.
  • the optimized driver model can be configured to the vehicle, so that the vehicle can determine whether there is abnormal driving behavior based on the optimized driver model; for another example, when determining that the first confidence level is greater than or equal to After the second threshold, the vehicle can optimize the driver model based on the first control parameter sequence and the corresponding multiple environmental information and motion state parameters, so that the driver model can better fit the user's control preferences and improve detection accuracy.
  • DMS driver monitoring system
  • the vehicle when the vehicle includes a driver monitoring system (DMS), when the user is driving the vehicle, due to the limitations of the user's clothing, camera shooting angle, blind spots and other factors, The DMS system may not be able to identify the user's fatigue driving behavior.
  • abnormal driving behavior can be identified based on the active control parameters and the driver model, so that it can When the DMS system fails to work properly, it can identify the user's abnormal driving behavior, which can help improve driving safety and reduce the occurrence of traffic accidents.
  • Embodiments of the present application also provide a device for implementing any of the above methods.
  • a device is provided that includes a unit for implementing each step performed by a chip, a vehicle, an abnormal driving behavior identification device, etc. in any of the above methods.
  • FIG. 6 is a schematic structural diagram of a device for identifying abnormal driving behavior provided by an embodiment of the present application.
  • the apparatus 700 may include an acquisition module 710 and a processing module 720.
  • the acquisition module 710 can be used to acquire the first control parameter, the first motion state parameter and the first environment information.
  • the first control parameter is the control parameter that controls the driving of the vehicle at the first moment.
  • the first motion state parameter The first environment information is used to indicate the movement state of the vehicle at the first moment, and the first environment information is used to indicate the surrounding environment of the vehicle at the first moment;
  • the processing module 720 may be used to, according to the first movement state parameters and the first environment information, based on the driver model, determine a first expected manipulation parameter; based on the first manipulation parameter and the first expected manipulation parameter, determine whether there is abnormal driving behavior.
  • the description of the first manipulation parameter, the first motion state parameter, and the first environment information may refer to step S410.
  • the description of the first manipulation parameter, the first motion state parameter, and the first environment information may refer to step S410.
  • details will not be repeated here.
  • the acquisition module 710 can also be used to obtain a historical manipulation parameter sequence
  • the processing module 720 can also be used to determine the driver model based on the historical manipulation parameter sequence.
  • the acquisition module 710 can also be used to: acquire the second manipulation parameter, the second motion state parameter and the second environment information.
  • the second manipulation parameter is the manipulation parameter that controls the driving of the vehicle at the second moment.
  • the second motion state parameter is used to indicate the motion state of the vehicle at the second moment, and the second environment information is used to indicate the surrounding environment of the vehicle at the second moment;
  • the processing module 720 can also be used to : According to the second motion state parameter and the second environment information, based on the driver model, determine a second expected manipulation parameter; the processing module 720 may be specifically configured to: based on the first manipulation parameter sequence and the first expected manipulation parameter A sequence to determine whether abnormal driving behavior exists, wherein the first operating parameter sequence includes the first operating parameter and the second expected operating parameter, and the first expected operating parameter sequence includes the first expected operating parameter and the second expected operating parameter. parameter.
  • steps S410 to S450 for description of the first manipulation parameter sequence and the first expected manipulation parameter sequence, reference may be made to steps S410 to S450, which will not be described again here for the sake of brevity.
  • the processing module 720 may also be configured to: determine a first confidence level based on the first manipulation parameter sequence and the first expected manipulation parameter sequence, the first confidence level being used to characterize the first manipulation parameter sequence and the first manipulation parameter sequence. The degree of similarity between the first expected manipulation parameter sequences; the processing module 720 is specifically configured to: when the first confidence is greater than or equal to the first threshold, based on the first operating parameter sequence and the first expected manipulation parameter sequence, Determine if there are any abnormal driving behaviors.
  • the description of the first confidence level may refer to step S450, which will not be described again here for the sake of brevity.
  • the processing module 720 may also be configured to: when the first confidence level is greater than or equal to the second threshold, calculate the first prediction model according to the first manipulation parameter sequence and the first expected manipulation parameter sequence. For optimization, the second threshold is smaller than the first threshold.
  • the acquisition module 710 can also be used to obtain the first information, which is used to indicate the user's steering preference; the processing module 720 can also be used to determine the driver's driving preference based on the first information. Model.
  • the processing module 720 can also be used to: determine an evaluation function according to the first information; the processing module 720 can be specifically used to: according to the first manipulation parameter sequence and the first expected manipulation parameter sequence, The value of the evaluation function is determined; when the value of the evaluation function is greater than or equal to the third threshold, it is determined that abnormal driving behavior exists.
  • the processing module 720 may also be used to: determine the operating condition type of the vehicle according to the first motion state parameter and/or the first state parameter; determine the operating condition type according to the vehicle operating condition type. Evaluation function.
  • the processing module 720 may also be used to: when it is determined that abnormal driving behavior exists, prompt the user that abnormal driving behavior exists.
  • the processing module 720 may also be used to control the vehicle to be in the automatic driving mode when it is determined that abnormal driving behavior exists.
  • the vehicle is a vehicle
  • the control parameters include: at least one of a steering wheel angle, an accelerator pedal stroke, and a brake pedal stroke.
  • the device for identifying abnormal driving behavior shown in Figure 6 can be used to implement the above method 400.
  • the device shown in Figure 6 can also be used to implement the method for identifying abnormal driving behavior described in method 500.
  • FIGS. 3 to 5 for the sake of simplicity, the embodiments of the present application will not describe them again.
  • FIG. 7 shows a schematic flowchart of a system for identifying abnormal driving behavior.
  • the system for identifying abnormal driving behavior may include a model determination unit, a residual calculation unit, an abnormal driving behavior identification unit, an abnormal driving behavior reminder unit, etc.
  • the model determination unit can be used to determine the parameters of an appropriate driver model based on historical driving data. For example, when the vehicle is powered on, after identifying the driver's identity, it can determine the parameters of the appropriate driver model based on the driver's history.
  • the control parameters determine the parameters of the corresponding driver model. For another example, the parameters of the corresponding driver model are determined according to the control preference selected by the driver. For the sake of simplicity, they will not be described again here.
  • the model determination unit can be located in the vehicle or in the cloud server. When the model determination unit is located in the cloud server, the vehicle can obtain the parameters of the driver model through network equipment, etc. It should be understood that this is not limited in the embodiments of the present application.
  • the residual calculation unit may be used to determine the residual vector according to the active control parameters and the expected control parameters output by the real driver, so as to determine the degree of similarity between the active control parameters and the expected control parameters.
  • the control parameters may include steering wheel angle, accelerator opening, brake pedal stroke, etc., which will not be described again here for the sake of brevity.
  • the abnormal driving behavior identification unit may be used to determine and identify abnormal driving behavior.
  • the abnormal driving behavior identification unit can also implement functions such as confidence calculation, evaluation function calculation, and working condition identification. The above functions can be combined to realize the identification of abnormal driving behavior. For the sake of brevity, no further details will be given here.
  • the abnormal driving behavior reminder unit can be used to remind abnormal driving behaviors.
  • relevant devices can be controlled to prompt abnormal driving behavior through text, sound, lights, etc. For the sake of brevity, no further details will be given here.
  • the division of the above-mentioned model determination unit, residual calculation unit, abnormal driving behavior detection unit, abnormal driving behavior reminder unit, etc. is only a logical division.
  • the above-mentioned processing module 720 may include the residual calculation unit, It may also include the abnormal driving behavior detection unit, etc.
  • the above-mentioned device 700 may include part of the model determination unit, the driver model, the residual calculation unit, the abnormal driving behavior detection unit and the abnormal driving behavior reminder unit. Or all of them, etc. For the sake of brevity, I won’t go into details here.
  • the driver controls the vehicle based on environmental information
  • the driver model can output expected steering parameters based on the vehicle's motion status information and environmental information, so that the residual calculation unit can obtain the driver's true output.
  • the abnormal driving behavior detection unit can determine whether there is abnormal driving behavior based on the residual sequence, such as identifying abnormal driving behavior when the first confidence level meets the preset conditions. When it is determined that abnormal driving behavior exists, the abnormal driving behavior reminder unit can prompt the existence of abnormal driving behavior.
  • the embodiment of the present application takes Indicates the steering wheel angle change rate, expressed as Indicates the change rate of the driving pedal stroke. Other symbols are similar. For the sake of simplicity, examples will not be given here.
  • FIG. 8 shows a schematic flow chart of optimizing a driver model.
  • the driver model may include a planning module and a control module.
  • the planning module may be used to plan the target speed, target path, etc. of the vehicle, and the control module may be used to determine expected manipulation parameters.
  • the driver model and abnormal driving behavior recognition unit can obtain the vehicle's motion status information, such as vehicle speed v, acceleration a, angular acceleration ⁇ , etc., and the residual calculation unit is based on the control parameters output by the user. and the control parameters output by the driver model, the residual vector can be determined, and the abnormal driving behavior identification unit can identify abnormal driving behavior based on the residual vector.
  • the driver model and abnormal driving behavior identification unit may also include a learning unit, which may be used to optimize or adjust the driver model, so that the driver model can more accurately identify the driver's abnormal driving. Behavior.
  • the learning module can be based on the objective function Based on the optimization algorithm (such as particle swarm optimization algorithm), the driver model is optimized, and the driver model meets the termination condition (for example, the number of iterations n reaches the preset threshold, and the change in the objective function corresponding to two consecutive iterations is less than error threshold), the optimization of the planning module and/or the control module of the driver model can be terminated; for another example, as shown in Figure 8, when no abnormal driving behavior is detected, or the duration of the abnormal driving behavior is less than or equal to When the time threshold T lim ′ is reached, the control parameters corresponding to the time period can be provided to the learning module, so that the learning module can optimize the driver model so that the output of the driver model is closer to the driver's driving behavior.
  • the optimization algorithm such as particle swarm optimization algorithm
  • FIG. 9 is a schematic flowchart of another method of identifying abnormal driving behavior provided by an embodiment of the present application.
  • the working condition of the vehicle can be determined based on environmental information, such as obstacle information, etc., and then the working condition under the working condition can be determined.
  • the working condition under the working condition can be determined.
  • the value of each weight coefficient (for example, at least one of k 1 , k 2 , k 3 ) in the evaluation function can also be used to determine the threshold J lim and the time threshold T lim of the evaluation function, so that it can be determined whether Abnormal driving behavior exists.
  • the online learning unit can optimize and adjust the driver model based on the recognition result of the abnormal driving behavior and the corresponding first control parameter, so that the expected control parameters output by the driver model unit can be Closer to the driver’s driving behavior.
  • FIGS. 7 to 9 is only an example for ease of explanation, and may also be combined with the content of FIGS. 2 to 6 , which is not limited by the embodiment of the present application.
  • each unit or module in the above device is only a division of logical functions. In actual implementation, it can be fully or partially integrated into a physical entity, or it can also be physically separated.
  • the units or modules in the device can be implemented in the form of the processor calling software; for example, the device includes a processor, the processor is connected to a memory, instructions are stored in the memory, and the processor calls the instructions stored in the memory to implement any of the above.
  • the processor is, for example, a general-purpose processor, such as a CPU or a microprocessor
  • the memory is a memory within the device or a memory outside the device.
  • the units in the device can be implemented in the form of hardware circuits, and some or all of the functions of the units can be implemented through the design of the hardware circuits, which can be understood as one or more processors; for example, in one implementation,
  • the hardware circuit is an ASIC, which realizes the functions of some or all of the above units through the design of the logical relationship of the components in the circuit; for another example, in another implementation, the hardware circuit can be implemented through PLD, taking FPGA as an example. It can include a large number of logic gate circuits, and the connection relationships between the logic gate circuits can be configured through configuration files to realize the functions of some or all of the above units. All units of the above device may be fully realized by the processor calling software, or may be fully realized by hardware circuits, or part of the units may be realized by the processor calling software, and the remaining part may be realized by hardware circuits.
  • the processor is a circuit with signal processing capabilities.
  • the processor may be a circuit with instruction reading and execution capabilities, such as a CPU, a microprocessor, a GPU, or DSP, etc.; in another implementation, the processor can implement certain functions through the logical relationship of the hardware circuit.
  • the logical relationship of the hardware circuit is fixed or reconfigurable.
  • the processor is an application-specific integrated circuit ASIC or programmable logic.
  • the process of the processor loading the configuration file and realizing the hardware circuit configuration can be understood as the process of the processor loading instructions to realize the functions of some or all of the above units.
  • it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as NPU, TPU, DPU, etc.
  • each unit in the above device can be one or more processors (or processing circuits) configured to implement the above method, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA , or a combination of at least two of these processor forms.
  • processors or processing circuits
  • each unit in the above device may be integrated together in whole or in part, or may be implemented independently. In one implementation, these units are integrated together and implemented as a system-on-a-chip (SOC).
  • SOC may include at least one processor for implementing any of the above methods or implementing the functions of each unit of the device.
  • the at least one processor may be of different types, such as a CPU and an FPGA, a CPU and an artificial intelligence processor, CPU and GPU etc.
  • FIG. 10 is an example structural diagram of a device 1300 provided by the embodiment of the present application.
  • Apparatus 1300 includes a processor 1302, a communication interface 1303, and a memory 1304.
  • One example of device 1300 is a chip.
  • Another example of apparatus 1300 is a computing device.
  • the processor 1302, the memory 1304 and the communication interface 1303 can communicate through a bus.
  • Executable code is stored in the memory 1304, and the processor 1302 reads the executable code in the memory 1304 to execute the corresponding method.
  • the memory 1304 may also include an operating system and other software modules required for running processes.
  • the executable code in the memory 1304 is used to implement the methods shown in FIGS. 3 to 5
  • the processor 1302 reads the executable code in the memory 1304 to execute the methods shown in FIGS. 3 to 5 .
  • the processor 1302 may be a CPU.
  • Memory 1304 may include volatile memory (VM), such as random access memory (RAM).
  • VM volatile memory
  • RAM random access memory
  • Memory 1304 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD) or solid-state boot ( solid state disk (SSD).
  • ROM read-only memory
  • HDD hard disk drive
  • SSD solid-state boot
  • An embodiment of the present application also provides a vehicle, which may include the above device 700, or the above device 1300, or a device for implementing methods 200 to 500.
  • Embodiments of the present application also provide a computer program product.
  • the computer program product includes: computer program code.
  • the computer program code When the computer program code is run on a computer, it causes the computer to execute the above method.
  • Embodiments of the present application also provide a computer-readable storage medium.
  • the computer-readable medium stores program code.
  • the computer program code When the computer program code is run on a computer, it causes the computer to execute the methods of FIG. 3 to FIG. 5 .
  • An embodiment of the present application also provides a chip, including: at least one processor and a memory.
  • the at least one processor is coupled to the memory and is used to read and execute instructions in the memory to execute the above-mentioned steps of Figures 3 to 5. method.
  • first”, second and other words are used to distinguish the same or similar items with basically the same functions and functions. It should be understood that the terms “first”, “second” and “nth” There is no logical or sequential dependency, and there is no limit on the number or execution order. For example, “first manipulation parameter” and “second manipulation parameter” are only used for distinction and do not mean that the “first manipulation parameter” and “second manipulation parameter” have different priorities.
  • the size of the sequence number of each process does not mean the order of execution.
  • the execution order of each process should be determined by its function and internal logic, and should not be used in the implementation of the embodiments of the present application.
  • the process constitutes any limitation.
  • determining B based on A does not mean determining B only based on A, and B can also be determined based on A and/or other information.
  • the size of the sequence numbers of the above-mentioned processes does not mean the order of execution.
  • the execution order of each process should be determined by its functions and internal logic, and should not be used in the embodiments of the present application.
  • the implementation process constitutes any limitation.
  • a component may be, but is not limited to, a process, a processor, an object, an executable file, a thread of execution, a program and/or a computer running on a processor.
  • applications running on the computing device and the computing device may be components.
  • One or more components can reside in a process and/or thread of execution and a component can be localized on one computer and/or distributed between 2 or more computers. Additionally, these components can execute from various computer-readable media having various data structures stored thereon.
  • a component may, for example, be based on a signal having one or more data packets (eg, data from two components interacting with another component, a local system, a distributed system, and/or a network, such as the Internet, which interacts with other systems via signals) Communicate through local and/or remote processes.
  • data packets eg, data from two components interacting with another component, a local system, a distributed system, and/or a network, such as the Internet, which interacts with other systems via signals
  • the disclosed systems, devices and methods can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components may be combined or can be integrated into another system, or some features can be ignored, or not implemented.
  • the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components shown 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 units may be selected according to the actual situation to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application can be integrated into one processing unit, each unit can exist physically alone, or two or more units can be integrated into one unit.
  • the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product.
  • the computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory, random access memory, magnetic disk or optical disk and other various media that can store program codes.

Abstract

一种异常驾驶行为识别的方法、装置，可以应用于新能源汽车或智能汽车中，能够在无需新增传感器的情况下，实现对异常驾驶行为的检测，也能够避免传感器在检测异常驾驶行为过程时所受到的限制。该方法包括：获取第一操纵参数、第一运动状态参数和第一环境信息（S410），第一操纵参数为在第一时刻控制交通工具行驶的操纵参数，第一运动状态参数用于指示交通工具在第一时刻的运动状态，第一环境信息用于指示交通工具在第一时刻所处的周边环境；根据第一运动状态参数和第一环境信息，基于驾驶员模型，确定第一预期操纵参数（S430）；根据第一操纵参数和第一预期操纵参数，确定是否存在异常驾驶行为（S450）。

Description

一种异常驾驶行为识别的方法、装置和交通工具 技术领域
本申请实施例涉及安全领域，更具体地，涉及一种异常驾驶行为识别的方法、装置和交通工具。
背景技术
近年来随着社会经济飞速发展，城市道路大规模建设，交通工具的数量高速增长，交通事故的发生率日益攀高。对于由驾驶员控制的交通工具而言，驾驶员在感知周围环境的条件下，根据自身驾驶经验控制该交通工具行驶，其行为对驾驶安全有至关重要的影响。而在驾驶过程中驾驶员难免会出现疲劳驾驶、注意力分散、情绪化驾驶、突发性疾病等情况，由此对交通工具的操纵行为，比如对车辆方向盘、油门踏板、制动踏板等的异常操作会直接影响到行驶安全。因此，对驾驶员的驾驶行为进行检测，并识别出其中的异常驾驶行为，对于提高驾驶员的驾驶能力、降低其驾驶负荷，以及从本质上减少交通事故的发生，具有重要的意义。
现有技术中对于异常驾驶行为的检测通常依靠摄像头采集用户身体或面部的图像，在进行特征提取后分析用户的异常驾驶行为，该方式在光线较差或用户被遮挡时无法进行检测；也有基于可穿戴传感器的检测方法，利用用户的生理特征数据分析并识别用户的异常驾驶状态，该方式又受到对用户体征数据采集的限制。而且，上述方法往往需要额外增加传感器的成本。因此如何独立、准确地对用户的异常驾驶行为进行检测，成为亟待解决的问题。
发明内容
本申请实施例提供一种异常驾驶行为识别的方法、装置，能够在无需增加额外传感器的情况下，独立、准确地对用户的异常驾驶行为进行识别，能够降低检测异常驾驶行为的成本。
第一方面，提供了一种异常驾驶行为识别的方法，包括：获取第一操纵参数、第一运动状态参数和第一环境信息，该第一操纵参数为在第一时刻控制交通工具行驶的操纵参数，该第一运动状态参数用于指示该交通工具在该第一时刻的运动状态，第一环境信息用于指示该交通工具在该第一时刻所处的周边环境；根据该第一运动状态参数和该第一环境信息，基于驾驶员模型，确定第一预期操纵参数；根据该第一操纵参数和该第一预期操纵参数，确定是否存在异常驾驶行为。
本申请实施例中，根据用户输出的操纵参数和驾驶员模型的预测操纵参数，确定是否存在对该交通工具的异常驾驶行为，使得可以在无需新增传感器的情况下，实现对异常驾驶行为的检测。而且由于该检测方式直接基于用户对于交通工具的操纵参数，可以避免传感器在检测异常驾驶行为时所受到的限制，也可以降低检测异常驾驶行为的延迟。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：获取历史操纵参数序列，该历史操纵参数序列包括，在该第一时刻前的第一历史时长内，控制该交通工具行驶的至少一个操纵参数；根据该历史操纵参数序列，确定该驾驶员模型。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：获取第二操纵参数、第二运动状态参数和第二环境信息，该第二操纵参数为在第二时刻控制该交通工具行驶的操纵参数，该第二运动状态参数用于指示该交通工具在该第二时刻的运动状态，第二环境信息用于指示该交通工具在该第二时刻所处的周边环境；根据该第二运动状态参数和该第二环境信息，基于该驾驶员模型，确定第二预期操纵参数；该根据该第一操纵参数和该第一预期操纵参数，确定是否存在异常驾驶行为，包括：根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，其中，该第一操作参数序列包括该第一操纵参数和该第二操纵参数，该第一预期操纵参数序列包括该第一预期操纵参数和该第二预期操纵参数。
由于用户对于交通工具的操纵通常具有连贯性，本申请实施例中，根据第一操纵参数序列和第一预期操纵参数序列，进行异常驾驶行为的识别，有助于提高对于识别异常驾驶行为的准确性。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：根据第一操纵参数序列和第一预期操纵参数序列，确定第一置信度，该第一置信度用于表征该第一操纵参数序列和该第一预期操纵参数序列间的相似程度；该根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，包括：在该第一置信度大于或等于第一阈值时，根据该第一操作参数序列和该第一预期操纵参数序列，确定是否存在异常驾驶行为。
本申请实施例中，在第一置信度大于或等于第一阈值时，根据第一操纵参数序列和第一预期操纵参数序列确定是否存在异常驾驶行为，可以避免由于环境信息、运动状态参数以及驾驶员模型运行错误等所导致的异常检测，可以提高异常驾驶行为检测的准确度。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：在该第一置信度大于或等于第二阈值时，根据该第一操纵参数序列和该第一预期操纵参数序列，对该第一预测模型进行优化，该第二阈值小于该第一阈值。
本申请实施例中，在第一置信度大于或等于第二阈值时，根据第一操纵参数序列对该驾驶员模型进行优化，可以使得该驾驶员模型所输出的操纵参数更加符合用户的驾驶习惯，从而可以提高异常驾驶行为检测的准确性。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：获取第一信息，该第一信息用于指示用户操纵偏好；根据该第一信息，确定该驾驶员模型。
本申请实施例中，由于通过根据用户操纵偏好确定该驾驶员模型，可以实现对于驾驶员模型的差异化配置，可以提高该异常驾驶行为检测的方法的准确度。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：根据该第一信息，确定评价函数；该根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，包括：根据该第一操纵参数序列和该第一预期操纵参数序列，确定该评价函数的值；在该评价函数的值大于或等于第三阈值时，确定存在异常驾驶行为。
本申请实施例中，根据第一信息确定评价函数，使得可以根据不同的用户操纵偏好，设定相应地异常驾驶行为的评价基准，从而可以降低误判的可能性，提高识别异常驾驶行 为的准确度。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：根据该第一运动状态参数和/或该第一状态参数，确定该交通工具的工况类型；根据该交通工具的工况类型，确定该评价函数。
本申请实施例中，通过确定该交通工具的工况类型，并由此确定评价函数，使得可以基于用户实际驾驶中的各种场景，设定相应的异常驾驶行为的评价基准，从而可以降低误判的可能性，提高识别异常驾驶行为的准确度。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：在确定存在异常驾驶行为时，提示用户存在异常驾驶行为。
本申请实施例中，在确定存在异常驾驶行为时，通过提示用户可以集中用户的注意力，从而可以减少交通事故的发生。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：在确定存在异常驾驶行为时，控制该交通工具处于自动驾驶模式。
结合第一方面，在第一方面的某些实现方式中，该交通工具为车辆，该操纵参数包括：方向盘转角、加速踏板行程和制动踏板行程中的至少一项。
第二方面，提供了一种异常驾驶行为识别的装置，该装置包括：获取模块，可以用于获取第一操纵参数、第一运动状态参数和第一环境信息，该第一操纵参数为在第一时刻控制交通工具行驶的操纵参数，该第一运动状态参数用于指示该交通工具在该第一时刻的运动状态，第一环境信息用于指示该交通工具在该第一时刻所处的周边环境；处理模块，可以用于根据该第一运动状态参数和该第一环境信息，基于驾驶员模型，确定第一预期操纵参数；根据该第一操纵参数和该第一预期操纵参数，确定是否存在异常驾驶行为。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：获取历史操纵参数序列，该历史操纵参数序列包括，在该第一时刻前的第一历史时长内，控制该交通工具行驶的至少一个操纵参数；根据该历史操纵参数序列，确定该驾驶员模型。
结合第二方面，在第二方面的某些实现方式中，获取模块，还可以用于：获取第二操纵参数、第二运动状态参数和第二环境信息，该第二操纵参数为在第二时刻控制该交通工具行驶的操纵参数，该第二运动状态参数用于指示该交通工具在该第二时刻的运动状态，第二环境信息用于指示该交通工具在该第二时刻所处的周边环境；该处理模块，还可以用于：根据该第二运动状态参数和该第二环境信息，基于该驾驶员模型，确定第二预期操纵参数；该处理模块，具体可以用于：根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，其中，该第一操作参数序列包括该第一操纵参数和该第二操纵参数，该第一预期操纵参数序列包括该第一预期操纵参数和该第二预期操纵参数。
结合第二方面，在第二方面的某些实现方式中，该处理模块，还可以用于：根据第一操纵参数序列和第一预期操纵参数序列，确定第一置信度，该第一置信度用于表征该第一操纵参数序列和该第一预期操纵参数序列间的相似程度；该处理模块，具体用于：在该第一置信度大于或等于第一阈值时，根据该第一操作参数序列和该第一预期操纵参数序列，确定是否存在异常驾驶行为。
结合第二方面，在第二方面的某些实现方式中，该处理模块，还可以用于：在该第一置信度大于或等于第二阈值时，根据该第一操纵参数序列，对该第一预测模型进行优化， 该第二阈值小于该第一阈值。
结合第二方面，在第二方面的某些实现方式中，获取模块，还可以用于获取第一信息，该第一信息用于指示用户操纵偏好；该处理模块，还可以用于根据该第一信息，确定驾驶员模型。
结合第二方面，在第二方面的某些实现方式中，该处理模块，还可以用于：根据该第一信息，确定评价函数；该处理模块，具体可以用于：根据该第一操纵参数序列和该第一预期操纵参数序列，确定该评价函数的值；在该评价函数的值大于或等于第三阈值时，确定存在异常驾驶行为。
结合第二方面，在第二方面的某些实现方式中，该处理模块，还可以用于：根据该第一运动状态参数和/或该第一状态参数，确定该交通工具的工况类型；根据该交通工具的工况类型，确定该评价函数。
结合第二方面，在第二方面的某些实现方式中，处理模块，还可以用于：在确定存在异常驾驶行为时，提示用户存在异常驾驶行为。
结合第二方面，在第二方面的某些实现方式中，处理模块，还可以用于：在确定存在异常驾驶行为时，控制该交通工具处于自动驾驶模式。
结合第二方面，在第二方面的某些实现方式中，该交通工具为车辆，该操纵参数包括：方向盘转角、加速踏板行程和制动踏板行程中的至少一项。
第三方面，提供了一种装置，该装置包括处理器和存储器，其中存储器用于存储程序指令，处理器用于调用该程序指令，以使该装置执行第一方面中任一种可能的方法。
第四方面，提供了一种交通工具，该交通工具包括第二方面或第三方面中任一项中的装置。
示例性地，该交通工具可以是智能汽车或新能源汽车等。
第五方面，提供了一种计算机程序产品，所述计算机程序产品包括：计算机程序代码，当该计算机程序代码在计算机上运行时，使得计算机执行上述第一方面中的方法。
第六方面，提供了一种计算机可读介质，所述计算机可读介质存储有程序代码，当所述计算机程序代码在计算机上运行时，使得计算机执行上述第一方面中的方法。
第七方面，提供了一种芯片，该芯片包括处理器与数据接口，该处理器通过该数据接口读取存储器上存储的指令，使得计算机执行上述第一方面中的方法。
附图说明
图1是本申请实施例提供的车辆的一个功能框图示意。
图2是本申请实施例提供的一种驾驶场景的示意图。
图3是本申请实施例提供的一种训练驾驶员模型的方法的示意性流程图。
图4是本申请实施例提供的一种异常驾驶行为识别的方法的示意性流程图。
图5是本申请实施例提供的另一种异常驾驶行为识别的方法的示意性流程图。
图6是本申请实施例提供的一种异常驾驶行为识别的装置的结构示意图。
图7是本申请实施例提供的一种识别异常驾驶行为的系统流程示意图。
图8是本申请实施例提供的一种优化驾驶员模型的流程示意图。
图9是本申请实施例提供的另一种识别异常驾驶行为的流程示意图。
图10为本申请实施例提供的一种装置的结构示例图。
具体实施方式
下面将结合附图，对本申请实施例中的技术方案进行描述。
下面将结合附图，对本申请实施例中的技术方案进行描述。
在本申请中，交通工具可以包括一种或多种不同类型的在陆地(例如，公路，道路，铁路等)，水面(例如：水路，江河，海洋等)或者空间上操作或移动的运输工具或者可移动物体。例如，交通工具可以包括汽车，自行车，摩托车，火车，地铁，飞机，船，飞行器，机器人或其它类型的运输工具或可移动物体等。以下以交通工具是车辆为例，简要介绍交通工具可以具有的功能。
示例性地，图1是本申请实施例提供的车辆100的一个功能框图示意。车辆100可以包括感知系统120、显示装置130和计算平台150，其中，感知系统120可以包括感测关于车辆100周边的环境的信息的若干种传感器。例如，感知系统120可以包括定位系统，定位系统可以是全球定位系统(global positioning system，GPS)，也可以是北斗系统或者其他定位系统、惯性测量单元(inertial measurement unit，IMU)、激光雷达、毫米波雷达、超声雷达以及摄像装置中的一种或者多种。
车辆100的部分或所有功能可以由计算平台150控制。计算平台150可包括处理器151至15n(n为正整数)，处理器是一种具有信号的处理能力的电路，在一种实现中，处理器可以是具有指令读取与运行能力的电路，例如中央处理单元(central processing unit，CPU)、微处理器、图形处理器(graphics processing unit，GPU)(可以理解为一种微处理器)、或数字信号处理器(digital signal processor，DSP)等；在另一种实现中，处理器可以通过硬件电路的逻辑关系实现一定功能，该硬件电路的逻辑关系是固定的或可以重构的，例如处理器为专用集成电路(application-specific integrated circuit，ASIC)或可编程逻辑器件(programmable logic device，PLD)实现的硬件电路，例如现场可编程门阵列(field programmable gate array，FPGA)。在可重构的硬件电路中，处理器加载配置文档，实现硬件电路配置的过程，可以理解为处理器加载指令，以实现以上部分或全部单元的功能的过程。此外，还可以是针对人工智能设计的硬件电路，其可以理解为一种ASIC，例如神经网络处理单元(neural network processing unit，NPU)、张量处理单元(tensor processing unit，TPU)、深度学习处理单元(deep learning processing unit，DPU)等。此外，计算平台150还可以包括存储器，存储器可以用于存储指令，处理器151至15n中的部分或全部处理器可以调用存储器中的指令，执行指令，以实现相应的功能。存储器还可用于存储数据，例如道路地图、路线信息，车辆的位置、方向、速度以及其它这样的车辆数据，以及其他信息。这种信息可在车辆100在自主、半自主和/或手动模式中操作期间被车辆100和计算平台150使用。
车辆100可以包括高级驾驶辅助系统(advanced driving assistant system，ADAS)，ADAS利用在车辆上的多种传感器(包括但不限于：激光雷达、毫米波雷达、摄像装置、超声波传感器、全球定位系统、惯性测量单元)从车辆周围获取信息，并对获取的信息进行分析和处理，实现例如障碍物感知、目标识别、车辆定位、路径规划、驾驶员监测/提醒等功能，从而提升车辆驾驶的安全性、自动化程度和舒适度。
应理解，图1中车辆的结构和功能，不应理解为对本申请实施例的限制。
可选地，其他交通工具可以具有与该车辆相似的功能，例如也包括感知系统，包括计算平台等，为了简洁，此处不再赘述。应理解，本申请实施例对此不做限定。
示例性地，图2是本申请实施例提供的一种驾驶场景的示意图。驾驶员通过感知环境信息，可以对车辆进行操纵，比如，该车辆可以包括车身系统、动力传动系统、底盘系统等，由此可以控制车辆加速、制动、转向等，又比如，该车辆可以是图1所示车辆100，该车辆可以通过其感知系统获知其所处环境信息，比如，可以获知所处环境的道路、道路围栏、交叉口等交通设施的信息；可以获知用于划分车道的路面的车道线、路面坑洼处、路面坡度等路面条件的信息；可以获取周边的行人、车辆等所在的位置、大小、速度等交通参与者的信息，可以获取阴天、晴天、雨天、雪天等天气条件的信息，等等，为了简洁，此处不再赘述。
示例性地，图3是本申请实施例提供的一种训练驾驶员模型的方法的示意性流程图，该方法200可以包括；
S210，获取第一操纵参数，所述第一操纵参数为该交通工具处于人工驾驶模式下的操纵参数。
示例性地，该第一操纵参数可以是控制交通工具行驶的操纵参数，也就是说，该第一操纵参数可以是用户控制该交通工具行驶所输出的操纵参数。例如，以该交通工具为车辆为例，该第一操纵参数可以是该车辆处于人工驾驶模式时，用户控制该交通工具的操纵参数，比如用户对于方向盘、加速踏板、制动踏板等可操纵工具的操纵，所输出的对应的方向盘转角、加速踏板行程、制动踏板行程等信息，对于传统燃油动力车辆，该第一操纵参数还可以包括比如离合器踏板行程等。应理解，以上关于第一操纵参数的方法只是举例，本申请实施例对此不做限定。
S220，获取交通工具的运动状态信息。该运动状态信息可以用于表征该交通工具的运动状态。
示例性地，该运动状态信息可以用于表征该交通工具的运动状态。例如，该运动状态信息可以包括交通工具的速度、加速度、角速度、角加速度等。示例性地，可以基于感知系统的传感器所采集的数据确定该交通工具的运动状态信息。例如，以交通工具为车辆为例，该车辆的感知系统中可以包括IMU，根据该IMU的采集结果，可以确定该车辆的加速度和/或角加速度；又例如，该车辆可以包括车速传感器，根据该车速传感器所采集的数据，可以确定该车辆的速度；再例如，该车辆还可以包括轮速传感器，基于该轮速传感器采集的数据可以确定车辆的速度。应理解，以上获取交通工具的运动状态信息的方式只是举例，以便于说明，本申请实施例对此不做限定。
S230，获取交通工具所处环境的环境信息。
示例性地，该环境信息可以用于描述该交通工具所处环境信息。例如，以交通工具为车辆为例，可以根据该车辆的感知系统所采集的数据，确定该车辆周围其他车辆、行人等交通元素的信息，根据该交通元素的信息，可以确定该车辆在行驶过程中可能存在的障碍物，比如可以获取周围其他车辆所处区域，周围车辆相对于该车辆的航向角、距离等，从而可以规划该车辆的行驶路径；可以根据车辆的传感器和/或车联网数据，确定车辆所在道路的信息，该道路信息可以是道路的斜度、倾角、道路的宽度等信息。应理解，根据该 环境信息可以确定该交通工具在行驶过程中所受到的环境的约束。应理解，以上关于环境信息的描述只是举例以便于说明，本申请实施例对此不做限定。
应理解，步骤S210至步骤S230中的部分或全部步骤，可以同时进行，也可以先进行步骤S210，也可以先进行S220，还可以先进行S230，本申请实施例对此不做限定。
S240，根据该第一操纵参数、运动状态信息和环境信息，进行模型训练，得到驾驶员模型，该驾驶员模型可以用于预测控制该交通工具的操纵参数。
示例性地，该第一操纵参数，与该运动状态参数和环境信息存在对应关系。例如，该第一操纵参数可以是在任一时刻控制该交通工具行驶的操纵参数，该第一运动状态信息可以指示该交通工具在该时刻的运动状态，该环境信息可以指示该交通工具在该时刻所处的周边环境，也就是说，在该时刻，用户根据该交通工具的运动状态和周边环境，对于该交通工具的控制可以体现为该第一操纵参数，因此，可以根据该第一操纵参数、对应的运动状态参数和环境信息，进行模型训练，从而可以得到用于预测控制该交通工具的操纵参数的驾驶员模型。也就是说，基于该训练得到的驾驶员模型，可以根据该交通工具对于实时的环境和其自身运动状态的感知，输出模拟自动驾驶的操纵参数。
示例性地，在实际中，由于各传感器的采样频率可能存在差异，所获取的运动状态参数和环境信息，从严格意义而言，可能并不指示该第一操纵参数所在时刻该交通工具的运动状态和周边环境，当该运动状态参数所对应的时刻、该环境信息所对应的时刻、该第一操纵参数所对应的时刻之间的时间差，在预设阈值之内时，可以认为该运动状态参数所对应的时刻、该环境信息所对应的时刻、该第一操纵参数所对应的时刻为同一时刻。例如，意在获取第10秒(second，s)的第一操纵参数、运动状态参数和环境信息，由于实际工作中采样频率等因素的影响，所获取的第一操纵参数可以是第10.05s时控制该交通工具行驶的操纵参数，所获取的运动状态参数可以是该交通工具在第9.91s的运动状态，所获取的对应的环境信息可以是该交通工具在第10.03s所处的周边环境，由于对应的时刻间的时间差小于或等于预设时间差阈值(比如0.1s)，可以认为该所获取运动状态参数和环境信息，所指示该交通工具在该第一操纵参数对应的时刻下的运动状态和周边环境，也就是说，该第一操纵参数、运动状态参数和周边环境所对应的时刻为同一时刻(比如第10s)，即，可以认为该获取的运动状态参数指示了该交通工具在第10s的运动状态，该环境信息指示了该交通工具在第10s的周边环境，可以认为该第一操纵参数为第10s时控制该交通工具行驶的操纵参数。
示例性地，可以是根据至少一组第一操纵参数，以及对应的至少一组运动状态信息、环境信息，进行模型训练，得到该驾驶员模型。例如，根据多组第一操纵参数，以及对应的运动状态信息、环境信息确定驾驶员模型，可以是基于优化算法，比如禁忌算法、遗传算法、粒子群优化算法、蚁群算法等确定驾驶员模型，从而可以获得与用户实际操纵行为相匹配的驾驶员模型。为了简洁此处不再赘述。
示例性地，对于不同驾驶偏好的用户，在相同的驾驶场景下，对于交通工具的操纵可以存在不同。例如，以交通工具为车辆为例，驾驶偏好激进的用户，以及驾驶偏号保守的用户，或称为，第一信息为“激进”以及“保守”所对应的用户，二者在面对相同的驾驶场景，比如在超车、换道等驾驶场景时，对于方向盘、加速踏板、制动踏板、档位等的操纵可能会存在较大的差异，从而在相同的场景中，二者所驾驶的车辆可以具有不同的运动 状态(比如速度、加速度等不同)，可以体现为二者所驾驶的车辆的可以具有不同的行驶轨迹等。
示例性地，由于不同驾驶偏好的用户，在相同的驾驶场景下，对于交通工具的操纵可能存在较大差异，也就是说，由不同驾驶偏好的用户所控制的交通工具，在相同的运动状态参数、环境信息的情况下，控制各交通工具行驶的第一操纵参数可能会存在较大差异，由此，根据不同驾驶偏好的用户，可以通过模型训练得到不同的驾驶员模型，从而可以根据不同的驾驶员模型对不同驾驶偏好的用户的驾驶行为进行预测，以提高预测的准确度。例如，以交通工具为车辆为例，比如，对于驾驶偏好保守的用户，可以采用预瞄式驾驶员模型，而对于驾驶偏好激进的用户，可以采用基于神经网络训练后得到的驾驶员模型，也就是说，对于不同驾驶风格的用户，可以使用不同类型的驾驶员模型；又例如，对于驾驶偏好保守的用户，可以采用基于神经网络训练后得到的驾驶员模型，而对于驾驶偏好激进的用户，也可以采用基于神经网络训练而得到的驾驶员模型，其中，上述两个驾驶员模型可以具有不同的特征参数，也就是说，对于不同驾驶风格的用户，可以使用同一类型而具有不同特征参数的驾驶员模型。
示例性地，在模型训练的过程中，以及在确定驾驶眼模型之后，还可以根据第一操纵参数、运动状态信息和环境信息对该驾驶员模型进行优化和调整，对于该驾驶员模型的优化和调整可以包括多种实现方式。在一些实现方式中，该过程可以包括多次迭代的过程。一次迭代的过程可以包括以下步骤：
S310，将运动状态信息、环境信息输入至驾驶员模型，通过驾驶员模型对该运动状态信息、环境信息进行处理，输出预测结果，也就是说，输出对该交通工具预测的操纵参数。
S320，根据该预测的操纵参数与第一操纵参数，可以通过损失函数计算第一损失值，该第一损失值可以表示预测结果与第一操纵参数之间的偏差，预测结果与第一操纵参数间的偏差越大，则第一损失值越大。
S330，根据第一损失值调整驾驶员模型。
以上示出了优化驾驶员模型中的一次迭代过程，当进行一次迭代后，可以检测当前是否已经满足优化中止条件，当不满足中止条件时，进行下一次迭代过程；当满足优化中止条件时，可以将本次迭代过程中所采用的驾驶员模型输出为最终用于预测驾驶员的操纵参数的驾驶员模型。
其中，该优化中止条件可以是迭代次数达到目标次数或者损失函数满足预设条件，还可以为在一定迭代次数内其能力没有提升。其中，该目标次数可以为预先设置的迭代次数，用以确定对于驾驶员模型的优化过程的结束时机，避免对训练资源的浪费；该终止条件可以是迭代过程中，损失函数值在一段时间内不变或者不下降，比如相邻两次或多次的损失函数的差的绝对值小于目标值，此时说明对于驾驶员模型的调整已经达到所需的效果。
应理解，以上优化驾驶员模型的方式只是举例，本申请实施例对此不做限定。
示例性地，可以由云服务器确定驾驶员模型，该云服务器还可以实现对于驾驶员模型的优化，该交通工具可以获取由云服务器确定/优化完成的驾驶员模型；也可以是在云服务器确定驾驶员模型后，交通工具获取该驾驶员模型，而且后续可以实现对该驾驶员模型的优化，还可以是由该交通工具确定该驾驶员模型，并实现对该驾驶员模型的优化，本申请实施例对此不做限定。
本申请实施例中，通过确定驾驶员模型，使得可以根据该驾驶员模型预测用户对交通工具的操纵参数，从而可以基于该预测的操纵参数，对异常驾驶行为进行检测。
示例性地，图4是本申请实施例提供的一种异常驾驶行为识别的方法的示意性流程图，该方法400包括以下步骤：
S410，获取第一操纵参数、第一运动状态参数和第一环境信息。
示例性地，该第一时刻可以为任一时刻，可以获取该时刻控制交通工具行驶的操纵参数，即第一操纵参数，也就是说，该第一操纵参数可以表征第一时刻用户对该交通工具的操纵，第一运动状态参数可以用于指示该交通工具在第一时刻的运动状态，该第一环境信息可以用于指示该交通工具在该第一时刻所处的周边环境。
为了便于理解和说明，控制交通工具行驶的操纵参数，或者说表征用户对于交通工具的控制的操纵参数，可以简称为主动操纵参数。
示例性地，由于用户可以在一定时间内持续对该交通工具进行驾驶，因此可以持续对用户的异常驾驶行为进行检测。例如，可以获取预设时长内的主动操纵参数，以及该时长内相应的运动状态信息和环境信息，以确定该时长内是否存在异常驾驶行为；又例如，为了节省所占用的资源，可以基于一定的采样频率，获取该预设时长内的多组主动操纵参数以及对应的运动状态信息和环境信息，该多组操纵参数可以按照时间顺序形成序列。
为了便于理解和说明，本申请实施例对多个时刻的主动操纵参数进行区分，示例性地，可以将第一时刻的主动操纵参数，定义为第一操纵参数，相应地，将用于指示交通工具在该时刻的运动状态的运动状态参数，定义为第一运动状态参数，将用于指示交通工具在该时刻的周边环境的环境信息，定义为第一环境信息，可以将第二时刻的主动操纵参数，定义为第二操纵参数，相应地，可以将该时刻的运动状态参数和环境信息，定义为第二运动状态参数和第二环境信息。在该第一时刻之前第一历史时长之内，用于控制车辆行驶的操纵参数，可以称为历史操纵参数，该第一历史时长之内的至少一个历史操纵参数所构成的序列，可以称为历史操纵参数序列，该历史操纵参数可以是本次行驶中控制车辆行驶的操纵参数，也可以是用户上一次驾车过程中控制该车辆行驶的操纵参数，该第一历史时长可以为任意时长，比如5分钟、10分钟等等。为了简洁，此处不再赘述。
示例性地，可以获取多个时刻的主动操纵参数，以及该多个时刻对应的运动状态参数和环境信息。示例性地，可以获取第一操纵参数序列、运动状态参数序列和环境信息序列，该第一操纵参数序列，可以包括多个时刻下的主动操纵参数，比如可以包括第一操纵参数和第二操纵参数，也可以包括其他时刻的主动操纵参数，该运动状态参数序列，可以包括该多个时刻下对应的运动状态参数，该环境信息序列，可以包括该多个时刻下对应的环境信息。例如，可以根据基于时间窗获取该时间窗内的多个主动操纵参数，该多个主动操纵参数可以以一定的顺序(比如时间顺序等)构成第一操纵参数序列，相应地，可以确定该时间窗内的运动状态参数序列和环境信息序列，该时间窗的时长可以是固定时长，也可以是根据驾驶场景所确定的时长等，本申请实施例对此不做限定。
应理解，以上关于获取第一操纵参数、第一运动状态参数和第一环境信息的描述只是示例，本申请实施例对此不做限定。
一些可能的实现方式中，可以获取第一信息，该第一信息可以用于指示用户操纵偏好。
示例性地，可以根据用户标识，确定该第一信息。例如，可以获取用户标识，比如可 以通过获取用户的声纹信息、面部特征、获取用户所使用的账户信息等方式，确定用户标识，根据用户标识，比如可以通过服务器在服务器查询、在本地数据库查询等方式，确定与该用户标识相关的用户操纵偏好，从而可以根据用户标识确定第一信息。应理解，获取用户标识的方法可以参照现有技术，为了简洁，此处不再赘述。
示例性地，可以根据历史操纵数据，确定该第一信息。例如，以交通工具为车辆为例，可以根据用户在一段时长内对于车辆的操纵，确定该用户的驾驶风格，比如，可以预设不同驾驶风格的用户在不同场景下的预期的驾驶路径、操纵数据等，可以根据用户在一定时长(比如5分钟)内对于车辆的操纵数据，与该预设的不同驾驶风格的预期的操纵数据进行匹配，从而可以确定该第一信息；又例如，同一用户在不同场景下的驾驶偏好也可能存在差异，在交通工具所处的工况发生变化后，可以根据用户在一定时长内对于车辆的操纵，确定第一信息是否发生变化，在第一信息发生变化时，相应地，可以确定驾驶员模型，从而可以确保该异常行为检测方法的准确性。
示例性地，可以根据用户信息，确定该第一信息。例如，以交通工具为车辆为例，在用户首次驾驶该车辆时，可以通过手机、平板电脑等终端，车辆中控屏等座舱屏幕设定其驾驶偏好；又例如，当用户在驾驶该车辆时，可以通过手机、平板电脑等终端，车辆中控屏等座舱屏幕输入其性别、年龄、驾驶年限等用户信息，可以根据该用户信息基于存储于本地或服务器的用户特性表，确定相匹配的第一信息，该用户特性表可以用于确定用户信息与第一信息间的映射关系，比如可以基于对不同年龄、性别、驾驶年限、受教育程度等不同用户的操纵偏好的分析，确定该用户特性表，为了简洁此处不再赘述。
应理解，以上关于获取第一信息的方式只是举例以便于说明，本申请实施例对此不做限定。
应理解，该第一信息，可以是任意的数据格式，例如数字、字母、字符串等。示例性地，可以采用不同的标签表示该操纵偏好。例如，可以采用“激进”、“保守”、“中立”等标签表示该第一信息，其中标签“激进”可以表示用户的操纵方式较激进，以交通工具为车辆为例，比如，在驾驶车辆至红绿灯路口，而在绿灯通行时间较短的时，操纵方式激进的用户可能会提高车速通过路口，相应地，操纵方式保守的用户可能会降低车速等待下个红绿灯。示例性地，可以采用数字、字母、字符表示不同的驾驶风格。例如，可以采用数字“1”、“2”、“3”等表示不同驾驶风格类型的用户，比如，以数字“1”表示操纵偏好激进的用户，以数字“5”表示操纵方式较保守的用户。应理解，以上关于驾驶风格的描述只是示例以便于说明，还可以以其他的方式对不同的操纵偏好进行区分，本申请实施例对此不做限定。
应理解，本申请中所涉及的用户信息的获得、保存、利用、处理等操作/行为，均为在符合当地法律规定下的合规使用。例如，本专利中所涉及的用户信息的获得、保存、利用、处理等操作表示在已获得个人同意情况下的操作。
一些可能的实现方式中，可以根据该第一信息，确定驾驶员模型。
示例性地，由于第一信息可以用于指示用户操纵偏好，可以根据该第一信息，确定相匹配的驾驶员模型。例如，根据该第一信息，比如以查询的方式等，确定对应相匹配的驾驶员模型，为了简洁，此处不再赘述。
由于不同的用户可能具有不同的操纵偏好，同一用户在不同工况可能具有不同的操纵 偏好，本申请实施例中，可以通过获取第一信息，从而可以基于该第一信息，确定驾驶员模型，由此可以提高异常驾驶行为检测的准确度。
S430，根据该第一运动状态参数和该第一环境信息，基于驾驶员模型，可以确定第一预期操纵参数。
示例性地，根据交通工具的运动状态参数和对应的环境信息，基于驾驶员模型，可以在该运动状态和周边环境下，预测用户对于该交通工具的操纵参数，即预期操纵参数，该预期操纵参数可以指基于该驾驶员模型所确定的操纵参数。例如，可以基于方法200所确定的驾驶员模型，根据第一时刻该交通工具的周边环境和运动状态，可以确定该时刻的预期操纵参数，比如，可以根据运动状态参数和环境信息，规划该交通工具在此场景下的目标速度和/或目标路径，从而根据该目标速度和/或目标路径，驾驶员模型可以输出预期操纵参数，以预测用户在交通工具处于该运动状态和所处环境时，对于该交通工具的操纵，即第一预期操纵参数。应理解，以上关于确定第一预期操纵参数的方法只是示例，以便于说明，本申请实施例对此不做限定。
示例性地，在获取多个时刻的用户输出的操纵参数，以及对应的多个运动状态参数和环境信息时，基于该驾驶员模型，可以确定多个对应的预期操纵参数。示例性地，在获取第一操纵参数序列时，相应地，根据对应的运动状态参数序列和环境信息序列，基于驾驶员模型，可以确定第一预期操纵参数序列，该第一预期操纵参数序列可以包括多个预期操纵参数。为了简洁，此处不再赘述。
一些可能的实现方式中，可以根据历史操纵参数序列，确定该驾驶员模型。例如，在车辆行驶过程中，可以实时获取并记录主动操纵参数，可以根据用户在一定时长内对于车辆的操纵，确定与之匹配的驾驶员模型，也可以根据该时间内的主动操纵参数对驾驶员模型进行优化、调整，从而可以确保该异常行为检测方法的准确性。
S450，根据该第一操纵参数和该第一预期操纵参数，确定是否存在异常驾驶行为。
示例性地，在该第一操纵参数和该第一预期操纵参数存在较大偏差时，可以确定存在异常驾驶行为。例如，以该交通工具为车辆，操纵参数为加速踏板行程、制动踏板行程和方向盘转角为例，在获取的第一操纵参数中，加速踏板行程为15厘米(centimeter，cm)、制动踏板行程为0、方向盘转角为0，即用户在该场景下并未控制车辆减速，而由驾驶员模型所确定的预期操纵参数中，加速踏板行程为0厘米、制动踏板行程为15厘米，方向盘转角为0，即驾驶员模型预测在该场景下用户需要控制车辆减速，由此，可以认为存在异常驾驶行为；又例如，可以通过查表的方式，确定第一操纵参数和第一预期操纵参数间的差值的阈值，二者间的差值大于或等于该阈值时，可以确定存在异常驾驶行为。应理解，以上确定异常驾驶行为的方式只是示例，以便于说明，本申请实施例对此不做限定。
示例性地，在获取多个主动操纵参数时，可以根据该多个主动操纵参数，以及基于驾驶员模型所确定的对应的多个预期操纵参数，确定是否存在异常驾驶行为。例如，可以根据该多个主动操纵参数和多个预期操纵参数，确定每个主动操纵参数和对应的预期操纵参数间的偏差，在该多个主动操纵参数与该多个预期操纵参数间的偏差的累积值，大于或等于设定阈值时，可以认为存在异常驾驶行为，该设定阈值可以是提前预设的，也可以是根据该交通工具所处的工况所计算得到的，本申请对此不做限定；又例如，可以基于评价函数，根据该多个主动操纵参数和预期操纵参数，确定该评价函数的值，由此可以确定是否 存在异常驾驶行为。
示例性地，以交通工具为车辆为例，比如，第一操纵参数可以包括用户输出的方向盘转角、加速踏板行程、制动踏板行程等，相应地，第一预期操纵参数也可以包括驾驶员模型所确定的方向盘转角、加速踏板行程、制动踏板行程等。为了便于说明，本申请实施例后续，以(δ、α、F b)表示该第一操纵参数，其中，δ可以用于表示用户输出的方向盘转角，α可以用于表示用户输出的加速踏板行程，F b可以用于表示用户输出的制动踏板行程，该加速踏板行程和制动踏板行程可以是实际的踏板行程，也可以是归一化后的踏板行程(比如以踏板总的行程对用户输出的踏板行程进行归一化等)，相应地，以(δ m、α m、F bm)表示该第一预期操纵参数，其中，δ m可以用于表示驾驶员模型所确定的方向盘转角，α m可以用于表示驾驶员模型所确定的加速踏板行程，F bm可以用于表示驾驶员模型所确定的制动踏板行程。应理解，第一操纵参数和第一预期操纵参数还可以是车辆档位等其他操纵参数，为了简洁此处不再一一举例说明，应理解，以上对于第一操纵参数和第一预期操纵参数的表征方式只是举例，以便于说明，本申请实施例对此不做限定。
示例性地，根据第一操纵参数和第一预期操纵参数，可以确定第一置信度，该第一置信度可以用于表示第一操纵参数与第一预期操纵参数间的相似程度。例如，以交通工具为车辆为例，根据第一操纵参数(δ、α、F b)和第一预期操纵参数(δ m、α m、F bm)，可以确定二者间的残差向量
Figure PCTCN2022095804-appb-000001
其中，
Figure PCTCN2022095804-appb-000002
ε 1＝(δ-δ m)，ε 2＝(α-α m)，ε 3＝(F b-F bm)，也就是说，该残差向量中的元素可以用于表示用户所输出的多个操纵参数与驾驶员模型所预测的对应的多个操纵参数之间的差异，相应地，该残差向量的长度
Figure PCTCN2022095804-appb-000003
可以用于表示第一操纵参数和第一预期操纵参数间的差异程度，该残差向量的长度越大，可以表示该第一操纵参数和第一预期操纵参数间的差异越大，其中，
Figure PCTCN2022095804-appb-000004
一些可能的实现方式中，为便于表征和统计，可以将该残差向量的长度映射至区间[0,1]，以表征第一操纵参数与第一预期操纵参数间的相似程度，也就是说，第一置信度可以是该残差向量长度的函数。比如，以γ表示第一置信度，
Figure PCTCN2022095804-appb-000005
该第一置信度γ越大，可以表示该第一操纵参数和第一预期操纵参数间的相似程度越高。
应理解，以上确定第一置信度的方式只是举例以便于说明，还可以采用其他方式确定第一置信度，比如可以根据方向盘转角、加速踏板行程等操纵参数对驾驶安全的影响对多个操纵参数设定不同的权重，基于该权重确定第一置信度。应理解，本申请实施例对于确定第一置信度的方式不作限定。
示例性地，在第一置信度大于或等于第一阈值时，可以根据该第一操纵参数和第一预期操纵参数确定是否存在异常驾驶行为。
本申请实施例中，在第一置信度大于或等于第一阈值时，第一操纵参数和第一预期操纵参数间的符合性较好，可以认为驾驶员模型处于正常运行状态，由此根据第一操纵参数和第一预期操纵参数确定是否存在异常驾驶行为，可以降低对于异常驾驶行为的误判。
示例性地，可以根据第一时长内的多个主动操纵参数和多个预期操纵参数，确定是否存在异常驾驶行为。其中，该第一时长可以是主动操纵参数的采样时间窗的时长，也可以是用于检测异常驾驶行为所使用的时间窗的时长，该多个第一操纵参数和多个第一预期操纵参数一一对应。
本申请实施例中，根据第一时长内的多个主动操纵参数和对应的预期操纵参数，可以 检测该时间段内的异常驾驶行为。由于用户对于交通工具的操纵通常具有连贯性，因此基于该第一时长内用户输出的多个操纵参数进行异常驾驶行为的检测和识别，有助于提高对于检测和识别异常驾驶行为的准确性。
示例性地，在用户对交通工具的操纵过程中，可以基于时间窗对其异常驾驶行为进行检测，该时间窗的时长可以是该第一时长，在该第一时长内，用户可以输出多组主动操纵参数，相应地，驾驶员模型也可以输出多组预期操纵参数。示例性地，该多组主动操纵参数可以以预设的排序方式构成第一操纵参数序列，相应地，该多组对应的预测操纵参数可以构成第一预期操纵参数序列，该第一操纵参数序列和该第一预期操纵参数序列可以具有相同的排序方式，也就是说，该第一操纵参数序列可以包括N个时刻的主动操纵参数，该第一预期操纵参数序列可以包括N个时刻预期操纵参数，该第一操纵参数序列中的主动操纵参数，与该第一预期操纵参数序列中的同一排序的预期操纵参数，可以存在一一对应关系，N为大于等于2的正整数。例如，以交通工具为车辆为例，比如，可以用(δ i、α i、F bi)表示该第一操纵参数序列中的第i组主动操纵参数(比如，可以为第一时刻的主动操纵参数)，相应地，可以使用(δ mi、α mi、F bmi)表示第一预期操纵参数序列中第i组的预期操纵参数(比如，可以为基于驾驶员模型所预测的该第一时刻的预期操纵参数)，相应地，可以使用
Figure PCTCN2022095804-appb-000006
表示该第一操纵参数序列中第i组主动操纵参数与其对应的预期操纵参数间的残差向量，根据多组主动操纵参数的顺序所组成的多个残差向量的序列，可以简称为残差序列，其中，i为小于或等于N的正整数。
示例性地，第一置信度也可以用于表示该多组主动操纵参数和多组预期操纵参数间的相似程度。例如，以交通工具为车辆为例，该第一时长内，可以包括n组主动操纵参数(δ i、α i、F bi)和预期操纵参数(δ mi、α mi、F bmi)，相应地，可以确定残差向量
Figure PCTCN2022095804-appb-000007
Figure PCTCN2022095804-appb-000008
其中，n为大于等于2的正整数，i为小于等于n的正整数，比如，可以根据欧式距离确定该多组主动操纵参数及多组预期操纵参数间的相似程度，为便于说明，以d表示该欧式距离，其中，
Figure PCTCN2022095804-appb-000009
第一置信度γ可以是该欧式距离d的函数。示例性地，为了便于表征和统计，可以将该欧式距离映射至区间[0,1]，以表征该多组主动操纵参数和多组预期操纵参数间的相似程度，即确定第一置信度，为了简洁，此处不再赘述。
应理解，以上关于第一操纵参数序列的描述，以及第一置信度的方法只是举例以便于说明，本申请实施例对此不做限定。
示例性地，在第一置信度大于或等于第一阈值时，可以根据该第一时长内的多组第一操纵参数和多组第二参数确定是否存在异常驾驶行为。
本申请实施例中，在第一置信度大于或等于第一阈值时，根据第一操纵参数序列和第一预期操纵参数序列确定是否存在异常驾驶行为，可以避免由于环境信息、运动状态参数以及驾驶员模型运行错误等所导致的异常检测，可以提高异常驾驶行为检测的准确度。
一些可能的实现方式中，在第一置信度大于或等于第二阈值时，可以根据该第一操纵参数序列优化驾驶员模型，该第二阈值可以小于或等于第一阈值。例如，在确定无异常驾驶行为，且该第一置信度大于或等于该第二阈值时，可以根据该第一操纵参数序列优化该驾驶员模型，可以使得该驾驶员模型所确定的预期操纵参数更加符合用户的驾驶行为，从 而可以提高异常驾驶行为检测的准确性。
示例性地，根据第一操纵参数序列和第一预期操纵参数序列确定是否存在异常驾驶行为，可以是根据残差序列，基于评价函数，确定是否存在异常驾驶行为。例如，以交通工具为车辆为例，该主动操纵参数和预期操纵参数可以是随时间变化的函数，比如，主动操纵参数(δ(t)、α(t)、F b(t))，预期操纵参数(δ m(t)、α m(t)、F bm(t))，时间窗的起止时间分别以t 1、t 2来表示，该残差向量也可以是随时间变化的函数，可以基于评价函数，确定是否存在异常驾驶行为，比如，评价函数可以为
Figure PCTCN2022095804-appb-000010
其中k 1、k 2、k 3可以是各操纵参数的权重系数，J可以是该评价函数的值，或称为第一评价值，该第一评价值可以用于评价是否存在异常驾驶行为，ε 1(t)可以用于表示用户输出的方向盘转角与驾驶员模型输出的方向盘转角间的差异随时间的变化，也就是说，ε 1(t)＝δ(t)-δ m(t)，当评价函数的计算结果，或者说第一评价值，大于或等于第三阈值时，可以确定车辆存在异常驾驶行为；又例如，在根据某一采样频率确定时间窗内的多个主动操纵参数时，可以以离散的方式计算该评价函数，为了简洁此处不再赘述。
示例性地，根据运动状态信息、环境信息中的至少一项，可以确定交通工具所处的工况。例如，以交通工具为车辆为例，可以根据车辆的运动状态(比如车辆的速度、加速度、角速度等)，环境信息(比如道路坡度、该道路下的期望车速等)和第一操纵参数(比如用户输出的方向盘转角、加速踏板行程、制动踏板行程等)中的一项或多项，可以通过片段划分、特征提取等处理方式，基于工况识别模型确定车辆所处的工况，比如根据上述参数确定车辆处于起步、加速、减速、转弯、高速巡航、换道避撞、停车等工况，也可以是其他工况类型，为了简洁不再一一举例说明。进一步地，可以根据所确定的工况，确定评价函数，以确定是否存在异常驾驶行为。例如，在评价函数针对多个操纵参数设定对应的多个权重系数时，比如，该评价函数为J＝k 1ε 1 2+k 2ε 2 2+k 3ε 3 2，可以根据工况确定该一个或多个权重系数，比如，k 1、k 2、k 3中的至少一项，从而确定该评价函数；又例如，由于不同工况下用户的驾驶行为可能存在不同，不同工况下的评价函数可以是不同的函数形式，比如，该评价函数可以是一次函数、二次函数、幂函数、指数函数等，为了简洁，此处不再一一举例说明。应理解，以上根据交通工具所处的工况确定评价函数的方法只是举例，本申请实施例对此不做限定。
示例性地，可以根据第一信息，确定评价函数。例如，由于基于用户操纵偏好，可以确定驾驶员模型，相应地，可以根据该第一信息，确定评价函数，比如该评价函数为J＝k 1ε 1 2+k 2ε 2 2+k 3ε 3 2，根据该第一信息可以确定对应的一组权重系数k 1、k 2、k 3，从而可以根据不同的用户操纵偏好，设定相应地异常驾驶行为的评价基准，从而可以降低误判的可能性，提高识别异常驾驶行为的准确度。
示例性地，可以结合工况和第一信息，确定评价函数。例如，在确定该交通工具的工况以及第一信息后，比如可以根据二维查表的方式等，确定相应的评价函数，为了简洁此处不再赘述。
示例性地，可以根据第一时长内所获取的多组第一操纵参数和第二操纵参数，基于评价函数，确定是否存在异常驾驶行为。例如，以交通工具为车辆，采样时间窗的时长为第一时长，其起止时间分别为t 1、t 2为例，评价函数可以为
Figure PCTCN2022095804-appb-000011
Figure PCTCN2022095804-appb-000012
可以根据该时间窗内的多组第一操纵参数和第二操纵参数间的残差，基于该 评价函数，确定是否存在异常驾驶行为，比如，在评价函数的计算结果大于或等于第三阈值时，可以确定存在异常驾驶行为；又例如，在第二时长内评价函数的计算结果大于或等于第三阈值时，可以确定存在异常驾驶行为，其中第二时长内可以包括一个或多个采样时间窗。应理解，以上确定异常驾驶行为的方法只是举例，本申请实施例对此不做限定。
应理解，以上关于评价函数的描述只是示例以便于说明，本申请实施例对此不做限定。
本申请实施例中，根据用户输出的操纵参数和驾驶员模型的操纵参数，确定是否存在对该交通工具的异常驾驶行为，使得可以在无需新增传感器的情况下，实现对异常驾驶行为的检测。而且由于该检测方式直接基于用户对于交通工具的操纵参数，可以避免传感器在检测异常驾驶行为时所受到的限制，也可以降低检测异常驾驶行为的延迟。进一步地，由于通过根据用户操纵偏好确定该驾驶员模型，可以实现对于驾驶员模型的差异化配置，可以提高该异常驾驶行为检测的方法的准确度。
在进行异常驾驶行为检测时，基于查询数据库的方式获取用户在正常操纵该交通工具时所输出的操纵参数，可能会需要庞大的数据库，对于数据库获取也可能会需要庞大的工程量，而且该数据库位于云端可能会收到通信质量的影响，由于存储容量的限制，存储于本地的数据库可能无法获取准确的操纵参数。由此，本申请实施例中，可以将预期操纵参数视作用户正常操纵该交通工具时所输出的操纵参数，基于驾驶员模型获取该预期操纵参数，可以避免存储空间、通信质量、数据库获取等方面的限制，而在用户对于交通工具的操纵过程中，由于周边环境、车辆运动状态、用户自身的操纵习惯等因素的影响，采用固定的驾驶员模型，会导致检测异常驾驶行为的准确度较低，本申请实施例中，通过历史操纵序列确定该驾驶员模型，可以提高识别异常驾驶行为的准确度，另外，基于用户的操纵偏好确定驾驶员模型，也可以实现对于驾驶员模型的差异化设置，也可以提高识别异常驾驶行为的准确度。
示例性地，以交通工具为车辆为例，图5是本申请实施例提供的一种异常驾驶行为识别的方法的示意性流程图，该方法500包括以下部分或全部步骤：
S510，获取第一操纵参数序列。
示例性地，在用户根据其对环境的感知控制车辆行驶时，可以获取控制该车辆行驶的主动操纵参数。例如，以车辆为车辆为例，该第一操纵参数可以包括方向盘转角、加速踏板行程、制动踏板行程等，车辆根据该第一操纵参数可以调整其速度、加速度、角速度等运动状态。
示例性地，可以获取用户在第一时长内输出的多个主动操纵参数。进一步地，该第一时长内的多个主动操纵参数可以构成第一操纵参数序列。为了简洁，此处不再赘述。
示例性地，第一操纵参数序列中所包括的主动操纵参数的数量可以大于或等于预设阈值。例如，以该预设阈值为6为例，可以由6个或6个以上的主动操纵参数构成该第一操纵参数序列，为了简洁，此处不再赘述。该预设阈值也可以是其他数值(比如3、10等)，本申请实施例对此不做限定。
S513，获取该车辆的运动状态信息和环境信息。
示例性地，关于车辆的运动状态信息和环境信息的描述可以参照步骤S220和S230，为了简洁，此处不再赘述。
示例性地，在获取多个主动操纵参数时，相应地，可以获取对应的多个运动状态参数 和环境信息。
应理解，获取主动操纵参数、车辆的运动状态信息和环境信息，可以同时获取，也可以是先获取主动操纵参数，还可以是先获取运动状态信息，还可以是先获取环境信息，本申请实施例对此不做限定。
S516，获取第一信息，根据第一信息确定驾驶员模型。
示例性地，可以根据历史操纵参数，确定该第一信息。例如，该历史操纵数据可以是此前保存在数据库中的主动操纵参数，比如，用户在前一次驾驶该交通工具时所保留的驾驶数据；该历史操纵参数，也可以是本次驾驶过程中所生成的操纵参数，比如，在本次驾驶中，用户已操纵该交通工具行驶15分钟，可以获取本次驾驶中10分钟内的主动操纵参数，比如可以通过对比、匹配等方式确定该用户的操纵偏好，从而确定该第一信息，为了简洁，此处不再赘述。
示例性地，根据第一信息可以确定驾驶员模型。
应理解，关于第一信息和驾驶员模型的描述，可以参考步骤S410至S430等，为了简洁，此处不再赘述。
S520，根据该车辆的运动状态信息和环境信息，基于驾驶员模型，确定第一预期操纵参数序列。
示例性地，关于确定预期操纵参数的描述可以操纵步骤S430，为了简洁，此处不再赘述。
示例性地，当获取第一时长内的多个主动操纵参数时，可以根据多组环境信息和车辆的运动状态信息，基于驾驶员模型，确定多个预期操纵参数，该多组预期操纵参数可以构成第一预期操纵参数序列。为了简洁，此处不再赘述。
S525，根据第一操纵参数序列和第一预期操纵参数序列，确定残差序列。
示例性地，根据多个主动操纵参数和对应的多个预期操纵参数，可以确定多个残差向量，该残差向量可以用于表示该一组第一操纵参数与对应的第二操纵参数间的差异，相应地，可以基于第一操纵参数序列的排序，将该多个残差向量构成残差序列，为了简洁此处不再赘述。
S530，确定第一置信度。
示例性地，可以根据残差序列确定第一置信度。例如，该多组主动操纵参数(δ i、α i、F bi)和预期操纵参数(δ mi、α mi、F bmi)的残差序列，可以以
Figure PCTCN2022095804-appb-000013
表示，比如，可以根据残差向量与零点间的欧式距离
Figure PCTCN2022095804-appb-000014
表征主动操纵参数和预期操纵参数间的相似程度，进一步地，为便于统计，可以将该欧式距离映射至区间[0,1]，将该映射结果作为第一置信度γ，比如，γ＝1-2*arctan(d)/π。应理解，以上确定第一置信度的方法只是举例以便于说明，本申请实施例对此不做限定。
S535，确定第一置信度是否大于或等于第一阈值，若第一置信度大于第一阈值，跳转步骤S545。
示例性地，关于第一阈值的描述可以参照步骤S450，为了简洁，此处不再赘述。
可选地，S540，根据第一操纵参数、环境信息和该车辆的运动状态信息，确定该车辆的工况。例如，根据该车辆的速度和加速度，可以确定该车辆是否处于加速的工况；根据 环境信息可以确定该车辆周围的障碍物分布情况，结合车辆的运动状态信息，可以确定该车辆是否处于换道避障的工况，等等。应理解，关于交通工具的工况、工况类型的描述可以参照相关技术，为了简洁，此处不再一一举例说明。
S545，确定评价函数。
示例性地，根据工况可以确定评价函数。例如，可以根据工况，可以以查表的方式确定该评价函数的形式，也可以根据工况确定评价函数中多个操纵参数的权重，为了简洁，此处不再一一举例说明。
示例性地，可以根据第一信息确定评价函数，也可以根据第一信息结合工况确定评价函数，为了简洁，此处不再赘述。
示例性地，关于评价函数的描述可以参照步骤S450，为了简洁此处不再赘述。
S550，确定是否存在异常驾驶行为，当存在异常驾驶行为时，可以跳转步骤S555。
示例性地，根据残差序列，可以通过评价函数确定第一评价值。例如，以采样时间窗的时长为第一时长，其起止时间分别为t 1、t 2，残差序列为
Figure PCTCN2022095804-appb-000015
评价函数为
Figure PCTCN2022095804-appb-000016
为例，根据工况和/或第一信息可以确定权重系数k 1、k 2、k 3中的至少一项，从而可以确定该评价函数，根据残差序列可以通过该评价函数确定第一评价值，在该第一评价值大于或等于第三阈值时，可以确定该采样时间窗内存在异常驾驶行为；又例如，在第二时长内可以包括多个第一时长的采样时间窗(比如，包括5个采样时间窗)，根据每个第一时长的采样时间窗的残差序列可以确定其对应的第一评价值，当该多个第一评价值大于或等于第三阈值的个数不低于预设数量(比如该预设数量为3)时，比如5个采样时间窗的第一评价值中有3个大于或等于第三阈值，可以确定该第二时长内存在异常驾驶行为；再例如，该第二时长内第一评价值大于或等于第三阈值的多个采样时间窗连续，且数量大于预设数量(比如该预设数量为2)时，可以确定该存在异常驾驶行为。应理解，以上确定是否存在异常驾驶行为的方法只是举例以便于说明，本申请实施例对此不做限定。
可选地，S555，提示用户存在异常驾驶行为。
示例性地，在确定存在异常驾驶行为时，可以提醒用户。例如，以交通工具为车辆为例，在确定存在异常驾驶行为时，可以在车辆中的屏幕显示文字、图像等提醒用户，比如，在车辆中控屏显示“请您注意驾驶安全”字样、通过抬头显示在车辆前风挡显示“请您安全驾驶”字样，也可以通过语音等提醒用户，比如，车辆通过蜂鸣器报警或者语音播报“请您注意驾驶安全”等，为了简洁不再一一举例说明。应理解，以上提醒用户的方式只是举例以便于说明，本申请实施例对此不做限定。
可选地，在确定存在异常驾驶行为时，可以用于触发第一功能。例如，该第一功能可以是自动驾驶功能、高级辅助驾驶功能等，比如，用户在人工驾驶过程中，可能由于视野盲区等的限制，用户对于交通工具的操纵可能被识别为异常驾驶行为，而由于车辆的感知系统的存在可以较好的实现对于周边环境的感知，在确定存在异常驾驶行为时，可以直接触发自动驾驶功能或高级辅助驾驶功能等，也可以通过语音提醒或中控屏，提示用户是否需要开启自动驾驶功能或高级辅助驾驶功能等，比如通过语音“建议您切换开启高级辅助驾驶”等；又例如，在车辆高速公路等场景驾驶时，在检测到异常驾驶行为之后，若在预设时间内持续检测到异常驾驶行为，可以触发自动泊车的功能，控制车辆在安全路段自动 泊车；又例如，在芯片、异常驾驶行为识别装置确定该存在异常驾驶行为时，可以是该芯片或异常行为识别装置等直接触发该第一功能，也可以是通过向其他装置发送消息，用于指示该其他装置触发该第一功能，为了简洁，此处不再赘述。
可选地，S560，确定第一置信度是否大于或等于第二阈值，若该第一置信度大于或等于第二阈值时，可以跳转步骤S565。
示例性地，在第一置信度大于或等于第二阈值时，可以根据该第一操纵参数序列对驾驶员模型进行优化。
示例性地，关于第二阈值的描述可以参照步骤S440，为了简洁此处不再赘述。
应理解，步骤S560和步骤S535可以同时进行，也可以先进行步骤S535，还可以先进行步骤S560，本申请实施例对此不做限定。
可选地，S565，驾驶员模型优化。
示例性地，在第一置信度大于或等于第二阈值时，可以根据该残差序列，基于优化算法优化驾驶员模型，比如，可以基于禁忌算法、遗传算法、粒子群优化算法、蚁群算法等，此处不再一一举例。例如，以粒子群优化算法为例，粒子在N维空间的位置可以分别代表需要优化的控制参数，可以采用模型预测控制方法建立/优化驾驶员模型，比如，X i＝(X 1,X 2,X 3,···,X n)＝(N p,N c,Q,R,Δv,Δy···)，其中，可以N p,N c,Q,R,Δv,Δy等可以分别代表预测时域、控制时域、控制矩阵参数、状态矩阵参数、预期速度修正量与预期横向位移修正量，也就是所需要优化的控制参数，比如，可以根据V id＝ωV id+C 1radm(0,1)(P id-X id)+C 1radm(0,1)(P gd-X id)，X id＝X id+V id更新各粒子速度和位置，其中C 1、C 2可以为加速因子，radm(0,1)可以为区间(0,1)之间的随机数，V id可以为第i个粒子在第d维上的速度矢量，X id可以表示第i个粒子在第d维上的位置矢量，ω为惯性因子，其值非负，P id为第i个粒子在迭代过程中的最优解的第d维变量的位置矢量，P gd为所有粒子中最优解的第d维变量的位置矢量，d为正整数。为了简洁此处不再赘述，应理解，关于粒子群优化算法的描述可以参照相关技术，为了简洁此处不再赘述。
示例性地，当该第一置信度大于或等于第二阈值时，可以根据该第一操纵参数序列和该第一预期操纵参数，通过损失函数优化该驾驶员模型。例如，可以根据第一操纵参数序列以及粒子群算法优化后的驾驶员模型所预测的操纵参数，确定残差序列，通过损失函数确定第一损失值，比如该损失函数可以是
Figure PCTCN2022095804-appb-000017
其中，cost可以为该第一损失值，f可以为该残差向第一损失值映射的函数关系式，即损失函数，从而根据该第一损失值调整该驾驶员模型。应理解，关于损失函数的描述可以参照相关技术，为了简洁此处不再赘述。
示例性地，可以由云服务器实现该驾驶员模型的优化，也可以是由车辆实现该驾驶员模型的优化。例如，在确定第一置信度大于或等于第二阈值后，可以将该第一操纵参数序列，以及对应的多个环境信息和运动状态参数发送至云服务器，从而云服务器可以实现对该驾驶员模型的优化，相应地，可以将优化后的驾驶员模型配置到车辆，从而该车辆可以基于该优化后的驾驶员模型确定是否存在异常驾驶行为；又例如，在确定第一置信度大于或等于第二阈值后，车辆可以根据该第一操纵参数序列，以及对应的多个环境信息和运动状态参数对驾驶员模型进行优化，从而使得该驾驶员模型可以更加贴合用户的操纵偏好，提高检测的准确性。
应理解，以上优化驾驶员模型的方法只是举例以便于说明，本申请实施例对此不做限定。
在实际驾驶中，由于用户对于交通工具的操纵具有连续性，通过获取一段时间内的第一操纵参数序列，并由此进行异常驾驶行为的检测，有助于提高对于检测异常驾驶行为的准确性。
示例性地，以交通工具为车辆为例，当车辆包括驾驶员监测系统(drivermonitorsystem，DMS)时，用户在驾驶该车辆的过程中，由于用户的着装、摄像头拍摄角度、盲区等因素的限制，该DMS系统可能无法识别到用户疲劳驾驶的行为，而由于用户对于交通工具的操纵均可以在操纵参数上体现，由此根据主动操纵参数，基于驾驶员模型，对异常驾驶行为进行识别，使得可以在该DMS系统无法正常工作的情况下，实现对于用户的异常驾驶行为的识别，从而可以有助于提高驾驶安全性，可以有助于减少交通事故的发生。
应理解，以上方法400可以和方法500相互结合，本申请实施例对此不作限定。
本申请实施例还提供用于实现以上任一种方法的装置，例如，提供一种装置包括用以实现以上任一种方法中芯片、车辆、异常驾驶行为识别装置等所执行的各步骤的单元。例如，请参考图6，其为本申请实施例提供的一种异常驾驶行为识别的装置的结构示意图。该装置700可以包括获取模块710和处理模块720。
其中，获取模块710，可以用于获取第一操纵参数、第一运动状态参数和第一环境信息，该第一操纵参数为在第一时刻控制交通工具行驶的操纵参数，该第一运动状态参数用于指示该交通工具在该第一时刻的运动状态，第一环境信息用于指示该交通工具在该第一时刻所处的周边环境；处理模块720，可以用于，根据该第一运动状态参数和该第一环境信息，基于该驾驶员模型，确定第一预期操纵参数；根据该第一操纵参数和该第一预期操纵参数，确定是否存在异常驾驶行为。
示例性地，关于第一操纵参数、第一运动状态参数和第一环境信息的描述可以参照步骤S410，为了简洁，此处不再赘述。
示例性地，该获取模块710，还可以用于获取历史操纵参数序列，该处理模块720，还可以用于根据该历史操纵参数序列，确定该驾驶员模型。
示例性地，获取模块710，还可以用于：获取第二操纵参数、第二运动状态参数和第二环境信息，该第二操纵参数为在第二时刻控制该交通工具行驶的操纵参数，该第二运动状态参数用于指示该交通工具在该第二时刻的运动状态，第二环境信息用于指示该交通工具在该第二时刻所处的周边环境；该处理模块720，还可以用于：根据该第二运动状态参数和该第二环境信息，基于该驾驶员模型，确定第二预期操纵参数；该处理模块720，具体可以用于：根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，其中，该第一操作参数序列包括该第一操纵参数和该第二操纵参数，该第一预期操纵参数序列包括该第一预期操纵参数和该第二预期操纵参数。
示例性地，关于第一操纵参数序列和第一预期操纵参数序列的描述，可以参照步骤S410至S450，为了简洁，此处不再赘述。
示例性地，该处理模块720，还可以用于：根据第一操纵参数序列和第一预期操纵参数序列，确定第一置信度，该第一置信度用于表征该第一操纵参数序列和该第一预期操纵参数序列间的相似程度；该处理模块720，具体用于：在该第一置信度大于或等于第一阈 值时，根据该第一操作参数序列和该第一预期操纵参数序列，确定是否存在异常驾驶行为。
示例性地，关于第一置信度的描述可以参照步骤S450，为了简洁此处不再赘述。
示例性地，该处理模块720，还可以用于：在该第一置信度大于或等于第二阈值时，根据该第一操纵参数序列和该第一预期操纵参数序列，对该第一预测模型进行优化，该第二阈值小于该第一阈值。
一些可能的实现方式中，该获取模块710，还可以用于获取第一信息，该第一信息用于指示用户操纵偏好；该处理模块720，还可以用于根据该第一信息，确定驾驶员模型。
示例性地，该处理模块720，还可以用于：根据该第一信息，确定评价函数；该处理模块720，具体可以用于：根据该第一操纵参数序列和该第一预期操纵参数序列，确定该评价函数的值；在该评价函数的值大于或等于第三阈值时，确定存在异常驾驶行为。
示例性地，该处理模块720，还可以用于：根据该第一运动状态参数和/或该第一状态参数，确定该交通工具的工况类型；根据该交通工具的工况类型，确定该评价函数。
示例性地，该处理模块720，还可以用于：在确定存在异常驾驶行为时，提示用户存在异常驾驶行为。
一些可能的实现方式中，该处理模块720，还可以用于：在确定存在异常驾驶行为时，控制该交通工具处于自动驾驶模式。
示例性地，该交通工具为车辆，该操纵参数包括：方向盘转角、加速踏板行程和制动踏板行程中的至少一项。
应理解，图6所示的异常驾驶行为识别的装置可以用于实现上述方法400，图6所示的装置还可以用于实现方法500所述的异常驾驶行为识别的方法，具体步骤可以参照上述对于图3至图5的描述，为了简洁，本申请实施例对此不再赘述。
示例性地，图7示出了一种识别异常驾驶行为的系统流程示意图。如图7所示，该用于识别异常驾驶行为的系统，可以包括模型确定单元、残差计算单元、异常驾驶行为识别单元、异常驾驶行为提醒单元等。
一些可能的实现方式中，模型确定单元，可以用于根据提供历史驾驶数据，确定适当的驾驶员模型的参数，比如，车辆上电时，在识别驾驶员身份后，可以根据该驾驶员的历史操纵参数，确定相应地的驾驶员模型的参数，又比如，根据驾驶员所选择的操纵偏好，确定相应的驾驶员模型的参数，为了简洁，此处不再赘述。该模型确定单元可以位于车辆，也可以位于云服务器，该模型确定单元位于云服务器时，车辆可以通过网络设备等获取该驾驶员模型的参数，应理解，本申请实施例对此不做限定。
示例性地，残差计算单元，可以用于根据真实驾驶员所输出的主动操纵参数和预期操纵参数确定残差向量，以便于确定主动操纵参数和预期操纵参数间的相似程度。例如，该操纵参数可以包括方向盘转角、油门开度、制动踏板行程等，为了简洁，此处不再赘述。
示例性地，异常驾驶行为识别单元，可以用于确定识别异常驾驶行为。例如，为了提高异常驾驶行为识别的准确度，该异常驾驶行为识别单元，也可以实现置信度计算、评价函数计算、工况识别等功能，可以结合上述功能实现异常驾驶行为的识别。为了简洁，此处不再赘述。
示例性地，异常驾驶行为提醒单元，可以用于对异常驾驶行为进行提醒。例如，可以控制相关装置通过文字、声音、灯光等方式提示存在异常驾驶行为。为了简洁，此处不再 赘述。
应理解，上述模型确定单元、残差计算单元、异常驾驶行为检测单元、异常驾驶行为提醒单元等的划分，仅仅是逻辑上的划分，比如，上述处理模块720，可以包括该残差计算单元，也可以包括该异常驾驶行为检测单元，等等，又比如，上述装置700，可以包括该模型确定单元、驾驶员模型、残差计算单元、异常驾驶行为检测单元和异常驾驶行为提醒单元中的部分或全部，等等，为了简洁，此处不再赘述。
一些可能的实现方式中，驾驶员根据环境信息，对车辆进行操纵，驾驶员模型可以根据车辆的运动状态信息和环境信息，输出预期操纵参数，由此残差计算单元可以获取该驾驶员真实输出的操纵参数，和该模型输出的操纵参数，从而可以确定残差序列。异常驾驶行为检测单元可以根据残差序列确定是否存在异常驾驶行为，比如在第一置信度满足预设条件时，进行异常驾驶行为识别等。在确定存在异常驾驶行为时，异常驾驶行为提醒单元，可以提示存在异常驾驶行为。
为了便于解释和说明，以操纵参数包括方向盘转角δ、加速踏板行程α、制动踏板行程F b为例，本申请实施例以
Figure PCTCN2022095804-appb-000018
表示方向盘转角变化率，以
Figure PCTCN2022095804-appb-000019
表示驾驶踏板行程变化率，其他符号类似，为了简洁，此处不再一一举例说明。
示例性地，图8示出了一种优化驾驶员模型的流程示意图。示例性地，如图8所示，驾驶员模型可以包括规划模块和控制模块，规划模块可以用于规划车辆的目标速度、目标路径等，控制模块可以用于确定预期操纵参数。
示例性地，如图8所示，驾驶员模型及异常驾驶行为识别单元，可以获取车辆的运动状态信息，比如车速v，加速度a，角加速度γ等，残差计算单元基于用户输出的操纵参数和驾驶员模型输出的操纵参数，可以确定残差向量，由此异常驾驶行为识别单元，可以根据该残差向量识别异常驾驶行为。进一步地，驾驶员模型及异常驾驶行为识别单元还可以包括学习单元，该学习单元，可以用于对驾驶员模型进行优化或调整，从而使得该驾驶员模型可以更加准确地识别驾驶员的异常驾驶行为。例如，该学习模块可以根据目标函数
Figure PCTCN2022095804-appb-000020
基于优化算法(比如粒子群优化算法)，对驾驶员模型进行优化，在该驾驶员模型满足终止条件(比如，迭代次数n达到预设阈值、连续两次迭代所对应的目标函数间的变化小于误差阈值)时，可以终止对驾驶员模型的规划模块和/或控制模块的优化；又例如，如图8所示，在未检测到异常驾驶行为，或者，异常驾驶行为的持续时间小于或等于时间阈值T lim′时，可以将该时间段对应的操纵参数提供给学习模块，从而该学习模块可以对驾驶员模型进行优化，使得该驾驶员模型的输出更接近于驾驶员的驾驶行为。
示例性地，图9是本申请实施例提供的另一种识别异常驾驶行为的流程示意图。
示例性地，如图9所示，根据环境信息，比如障碍物信息等可以确定车辆所处的工况，进而可以确定该工况下，根据车辆所处的工况，可以确定该工况下的评价函数中的各权重系数(比如，k 1、k 2、k 3中至少一项)的值，也可以用于确定评价函数的阈值J lim和时间阈值T lim，从而由此可以确定是否存在异常驾驶行为。相应地，在线学习单元可以根据该异常驾驶行为的识别结果，和相对应的第一操纵参数，对驾驶员模型进行优化和调整，从而可以使得该驾驶员模型单元所输出的预期操纵参数，能够更加接近于驾驶员的驾驶行为。
应理解，关于图7至图9的描述只是示例以便于说明，也可以与图2至图6的内容相结合，本申请实施例对此不做限定。
应理解，以上装置中各单元或模块的划分仅是一种逻辑功能的划分，实际实现时可以全部或部分集成到一个物理实体上，也可以物理上分开。此外，装置中的单元或模块可以以处理器调用软件的形式实现；例如装置包括处理器，处理器与存储器连接，存储器中存储有指令，处理器调用存储器中存储的指令，以实现以上任一种方法或实现该装置各单元的功能，其中处理器例如为通用处理器，例如CPU或微处理器，存储器为装置内的存储器或装置外的存储器。或者，装置中的单元可以以硬件电路的形式实现，可以通过对硬件电路的设计实现部分或全部单元的功能，该硬件电路可以理解为一个或多个处理器；例如，在一种实现中，该硬件电路为ASIC，通过对电路内元件逻辑关系的设计，实现以上部分或全部单元的功能；再如，在另一种实现中，该硬件电路为可以通过PLD实现，以FPGA为例，其可以包括大量逻辑门电路，通过配置文件来配置逻辑门电路之间的连接关系，从而实现以上部分或全部单元的功能。以上装置的所有单元可以全部通过处理器调用软件的形式实现，或全部通过硬件电路的形式实现，或部分通过处理器调用软件的形式实现，剩余部分通过硬件电路的形式实现。
在本申请实施例中，处理器是一种具有信号的处理能力的电路，在一种实现中，处理器可以是具有指令读取与运行能力的电路，例如CPU、微处理器、GPU、或DSP等；在另一种实现中，处理器可以通过硬件电路的逻辑关系实现一定功能，该硬件电路的逻辑关系是固定的或可以重构的，例如处理器为专用集成电路ASIC或可编程逻辑器件PLD实现的硬件电路，例如FPGA。在可重构的硬件电路中，处理器加载配置文档，实现硬件电路配置的过程，可以理解为处理器加载指令，以实现以上部分或全部单元的功能的过程。此外，还可以是针对人工智能设计的硬件电路，其可以理解为一种ASIC，例如NPU、TPU、DPU等。
可见，以上装置中的各单元可以是被配置成实施以上方法的一个或多个处理器(或处理电路)，例如：CPU、GPU、NPU、TPU、DPU、微处理器、DSP、ASIC、FPGA，或这些处理器形式中至少两种的组合。
此外，以上装置中的各单元可以全部或部分可以集成在一起，或者可以独立实现。在一种实现中，这些单元集成在一起，以片上系统(system-on-a-chip，SOC)的形式实现。该SOC中可以包括至少一个处理器，用于实现以上任一种方法或实现该装置各单元的功能，该至少一个处理器的种类可以不同，例如包括CPU和FPGA，CPU和人工智能处理器、CPU和GPU等。
示例性地，图10为本申请实施例提供的一种装置1300的结构示例图。装置1300包括处理器1302、通信接口1303和存储器1304。装置1300的一种示例为芯片。装置1300的另一种示例为计算设备。
处理器1302、存储器1304和通信接口1303之间可以通过总线通信。存储器1304中存储有可执行代码，处理器1302读取存储器1304中的可执行代码以执行对应的方法。存储器1304中还可以包括操作系统等其他运行进程所需的软件模块。
例如，存储器1304中的可执行代码用于实现图3至图5所示的方法，处理器1302读取存储器1304中的该可执行代码以执行图3至图5所示的方法。
其中，处理器1302可以为CPU。存储器1304可以包括易失性存储器(volatile memory，VM)，例如随机存取存储器(random access memory，RAM)。存储器1304还可以包 括非易失性存储器(non-volatile memory，NVM)，例如只读存储器(read-only memory，ROM)，快闪存储器，硬盘驱动器(hard disk drive，HDD)或固态启动器(solid state disk，SSD)。
本申请实施例还提供一种车辆，该车辆可以包括上述装置700，或者上述装置1300，或者用于实现方法200至500的装置。
本申请实施例还提供了一种计算机程序产品，该计算机程序产品包括：计算机程序代码，当该计算机程序代码在计算机上运行时，使得计算机执行上述方法。
本申请实施例还提供一种计算机可读存储介质，该计算机可读介质存储有程序代码，当该计算机程序代码在计算机上运行时，使得该计算机执行上述图3至图5的方法。
本申请实施例还提供一种芯片，包括：至少一个处理器和存储器，该至少一个处理器与该存储器耦合，用于读取并执行该存储器中的指令，以执行上述图3至图5的方法。
本申请中术语“至少一个”的含义是指一个或多个，本申请中术语“多个”的含义是指两个或两个以上。
本申请中术语“第一”“第二”等字样用于对作用和功能基本相同的相同项或相似项进行区分，应理解，“第一”、“第二”、“第n”之间不具有逻辑或时序上的依赖关系，也不对数量和执行顺序进行限定。例如，“第一操纵参数”和“第二操纵参数”仅用于区分，不代表“第一操纵参数”和“第二操纵参数”的优先级不同。
应理解，在本申请的各个实施例中，各个过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
应理解，根据A确定B并不意味着仅仅根据A确定B，还可以根据A和/或其它信息确定B。
应理解，本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。
应理解，在本申请的各种实施例中，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
在本说明书中使用的术语“部件”、“模块”、“系统”等用于表示计算机相关的实体、硬件、固件、硬件和软件的组合、软件、或执行中的软件。例如，部件可以是但不限于，在处理器上运行的进程、处理器、对象、可执行文件、执行线程、程序和/或计算机。通过图示，在计算设备上运行的应用和计算设备都可以是部件。一个或多个部件可驻留在进程和/或执行线程中，部件可位于一个计算机上和/或分布在2个或更多个计算机之间。此外，这些部件可从在上面存储有各种数据结构的各种计算机可读介质执行。部件可例如根据具有一个或多个数据分组(例如来自与本地系统、分布式系统和/或网络间的另一部件交互的二个部件的数据，例如通过信号与其它系统交互的互联网)的信号通过本地和/或远程进程来通信。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以 硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
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，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，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。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，可以是电性，机械或其它的形式。
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。可以根据实际情形选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器、随机存取存储器、磁碟或者光盘等各种可以存储程序代码的介质。
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以所述权利要求的保护范围为准。

Claims (27)

  1. 一种异常驾驶行为识别的方法，其特征在于，包括：
    获取第一操纵参数、第一运动状态参数和第一环境信息，所述第一操纵参数为在第一时刻控制交通工具行驶的操纵参数，所述第一运动状态参数用于指示所述交通工具在所述第一时刻的运动状态，所述第一环境信息用于指示所述交通工具在所述第一时刻所处的周边环境；
    根据所述第一运动状态参数和所述第一环境信息，基于驾驶员模型，确定第一预期操纵参数；
    根据所述第一操纵参数和所述第一预期操纵参数，确定是否存在异常驾驶行为。
  2. 如权利要求1所述的方法，其特征在于，所述方法还包括：
    获取历史操纵参数序列，其中，所述历史操纵参数序列包括，在所述第一时刻前的第一历史时长内，控制所述交通工具行驶的至少一个操纵参数；
    根据所述历史操纵参数序列，确定所述驾驶员模型。
  3. 如权利要求1或2所述的方法，其特征在于，所述方法还包括：
    获取第二操纵参数、第二运动状态参数和第二环境信息，所述第二操纵参数为在第二时刻控制所述交通工具行驶的操纵参数，所述第二运动状态参数用于指示所述交通工具在所述第二时刻的运动状态，第二环境信息用于指示所述交通工具在所述第二时刻所处的周边环境；
    根据所述第二运动状态参数和所述第二环境信息，基于所述驾驶员模型，确定第二预期操纵参数；
    所述根据所述第一操纵参数和所述第一预期操纵参数，确定是否存在异常驾驶行为，包括：
    根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，其中，所述第一操作参数序列包括所述第一操纵参数和所述第二操纵参数，所述第一预期操纵参数序列包括所述第一预期操纵参数和所述第二预期操纵参数。
  4. 如权利要求3所述的方法，其特征在于，所述方法还包括：
    根据第一操纵参数序列和第一预期操纵参数序列，确定第一置信度，所述第一置信度用于表征所述第一操纵参数序列和所述第一预期操纵参数序列间的相似程度；
    所述根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，包括：
    在所述第一置信度大于或等于第一阈值时，根据所述第一操作参数序列和所述第一预期操纵参数序列，确定是否存在异常驾驶行为。
  5. 如权利要求4所述的方法，其特征在于，所述方法还包括：
    在所述第一置信度大于或等于第二阈值时，根据所述第一操纵参数序列，对所述驾驶员模型进行优化，所述第二阈值小于所述第一阈值。
  6. 如权利要求1至5中任一项所述的方法，其特征在于，所述方法还包括：
    获取第一信息，所述第一信息用于指示用户操纵偏好；
    根据所述第一信息，确定所述驾驶员模型。
  7. 如权利要求6所述的方法，其特征在于，所述方法还包括：
    根据所述第一信息，确定评价函数；
    所述根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，包括：
    根据所述第一操纵参数序列和所述第一预期操纵参数序列，确定所述评价函数的值；
    在所述评价函数的值大于或等于第三阈值时，确定存在异常驾驶行为。
  8. 如权利要求1至7中任一项所述的方法，其特征在于，所述方法还包括：
    根据所述第一运动状态参数和/或所述第一运动状态参数，确定所述交通工具的工况类型；
    根据所述交通工具的工况类型，确定所述评价函数。
  9. 如权利要求1至8中任一项所述的方法，其特征在于，所述方法还包括：
    在确定存在异常驾驶行为时，提示用户存在异常驾驶行为。
  10. 如权利要求1至9中任一项所述的方法，其特征在于，所述方法还包括：
    在确定存在异常驾驶行为时，控制所述交通工具处于自动驾驶模式。
  11. 如权利要求1至10中任一项所述的方法，其特征在于，所述交通工具为车辆，所述操纵参数包括：方向盘转角、加速踏板行程和制动踏板行程中的至少一项。
  12. 一种异常驾驶行为识别的装置，其特征在于，包括：
    获取模块，用于获取第一操纵参数、第一运动状态参数和第一环境信息，所述第一操纵参数为在第一时刻控制交通工具行驶的操纵参数，所述第一运动状态参数用于指示所述交通工具在所述第一时刻的运动状态，第一环境信息用于指示所述交通工具在所述第一时刻所处的周边环境，所述交通工具处于人工驾驶模式；
    处理模块，用于根据所述第一运动状态参数和所述第一环境信息，基于驾驶员模型，确定第一预期操纵参数；根据所述第一操纵参数和所述第一预期操纵参数，确定是否存在异常驾驶行为。
  13. 如权利要求12所述的装置，其特征在于，所述获取模块还用于：
    获取历史操纵参数序列，其中，所述历史操纵参数序列包括，在所述第一时刻前的第一历史时长内，控制所述交通工具行驶的多组操纵参数；
    所述处理模块，还用于：
    根据所述历史操纵参数序列，确定所述驾驶员模型。
  14. 如权利要求12或13所述的装置，其特征在于，所述获取模块还用于：
    获取第二操纵参数、第二运动状态参数和第二环境信息，所述第二操纵参数为在第二时刻控制所述交通工具行驶的操纵参数，所述第二运动状态参数用于指示所述交通工具在所述第二时刻的运动状态，第二环境信息用于指示所述交通工具在所述第二时刻所处的周边环境；
    所述处理模块，还用于：根据所述第二运动状态参数和所述第二环境信息，基于所述驾驶员模型，确定第二预期操纵参数；
    所述处理模块，具体用于：
    根据第一操纵参数序列和第一预期操纵参数序列，确定是否存在异常驾驶行为，其中， 所述第一操作参数序列包括所述第一操纵参数和所述第二操纵参数，所述第一预期操纵参数序列包括所述第一预期操纵参数和所述第二预期操纵参数。
  15. 如权利要求14所述的装置，其特征在于，所述处理模块，还用于：
    根据第一操纵参数序列和第一预期操纵参数序列，确定第一置信度，所述第一置信度用于表征所述第一操纵参数序列和所述第一预期操纵参数序列间的相似程度；
    所述处理模块，具体用于：
    在所述第一置信度大于或等于第一阈值时，根据所述第一操作参数序列和所述第一预期操纵参数序列，确定是否存在异常驾驶行为。
  16. 如权利要求15所述的装置，其特征在于，所述处理模块，还用于：
    在所述第一置信度大于或等于第二阈值时，根据所述第一操纵参数序列，对所述驾驶员模型进行优化，所述第二阈值小于所述第一阈值。
  17. 如权利要求12至16中任一项所述的装置，其特征在于，所述获取模块还用于：
    获取第一信息，所述第一信息用于指示用户操纵偏好；
    所述处理模块，还用于根据所述第一信息，确定所述驾驶员模型。
  18. 如权利要求17所述的装置，其特征在于，所述处理模块，还用于：
    根据所述第一信息，确定评价函数；
    所述处理模块，具体用于：
    根据所述第一操纵参数序列和所述第一预期操纵参数序列，确定所述评价函数的值；
    在所述评价函数的值大于或等于第三阈值时，确定存在异常驾驶行为。
  19. 如权利要求12至18中任一项所述的装置，其特征在于，所述处理模块，还用于：
    根据所述第一运动状态参数和/或所述第一运动状态参数，确定所述交通工具的工况类型；
    根据所述交通工具的工况类型，确定所述评价函数。
  20. 如权利要求12至19中任一项所述的装置，其特征在于，所述处理模块，还用于：
    在确定存在异常驾驶行为时，提示用户存在异常驾驶行为。
  21. 如权利要求12至20中任一项所述的装置，其特征在于，所述处理模块，还用于：
    在确定存在异常驾驶行为时，控制所述交通工具处于自动驾驶模式。
  22. 如权利要求12至21中任一项所述的装置，其特征在于，所述交通工具为车辆，所述操纵参数包括：方向盘转角、加速踏板行程和制动踏板行程中的至少一项。
  23. 一种装置，其特征在于，包括处理器和存储器，所述存储器用于存储程序指令，所述处理器用于调用所述程序指令来执行权利要求1至11中任一项所述的方法。
  24. 一种交通工具，其特征在于，包括权利要求12至23中任一项所述的装置。
  25. 一种计算机可读存储介质，其特征在于，所述计算机可读介质存储有程序代码，当所述程序代码在计算机上运行时，使得计算机执行如权利要求1至11中任意一项所述的方法。
  26. 一种计算机程序产品，其特征在于，包括计算机程序代码，所述计算机程序代码在计算机上运行时，使得计算机执行如权利要求1至11中任意一项所述的方法。
  27. 一种芯片，其特征在于，所述芯片包括处理器与数据接口，所述处理器通过所述数据接口读取存储器上存储的指令，以执行如权利要求1至11中任一项所述的方法。
PCT/CN2022/095804 2022-05-28 2022-05-28 一种异常驾驶行为识别的方法、装置和交通工具 WO2023230740A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/095804 WO2023230740A1 (zh) 2022-05-28 2022-05-28 一种异常驾驶行为识别的方法、装置和交通工具

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/095804 WO2023230740A1 (zh) 2022-05-28 2022-05-28 一种异常驾驶行为识别的方法、装置和交通工具

Publications (1)

Publication Number Publication Date
WO2023230740A1 true WO2023230740A1 (zh) 2023-12-07

Family

ID=89026458

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/095804 WO2023230740A1 (zh) 2022-05-28 2022-05-28 一种异常驾驶行为识别的方法、装置和交通工具

Country Status (1)

Country Link
WO (1) WO2023230740A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117644870A (zh) * 2024-01-30 2024-03-05 吉林大学 一种基于情景感知的驾驶焦虑检测与车辆控制方法及系统

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008056059A (ja) * 2006-08-30 2008-03-13 Equos Research Co Ltd 運転者状態推定装置及び運転支援装置
CN109949438A (zh) * 2019-03-20 2019-06-28 深圳市赛梅斯凯科技有限公司 异常驾驶监测模型建立方法、装置及存储介质
US20210150827A1 (en) * 2019-11-19 2021-05-20 D.S. Raider Ltd System and method for monitoring and predicting breakdowns in vehicles
CN113119981A (zh) * 2021-04-09 2021-07-16 东风汽车集团股份有限公司 车辆主动安全控制方法、系统及存储介质
CN113291311A (zh) * 2021-05-28 2021-08-24 同济大学 一种紧急避撞工况下驾驶员异常行为检测方法及存储介质
CN113815631A (zh) * 2021-09-10 2021-12-21 华人运通(江苏)技术有限公司 基于历史驾驶行为的驾驶提醒方法、装置、设备及介质
CN113901979A (zh) * 2021-08-24 2022-01-07 浙江迪嘉智能技术有限公司 一种驾驶趋势预测方法及系统
CN113963533A (zh) * 2021-09-15 2022-01-21 上海钧正网络科技有限公司 驾驶行为异常检测方法、装置、电子设备、服务器及介质

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008056059A (ja) * 2006-08-30 2008-03-13 Equos Research Co Ltd 運転者状態推定装置及び運転支援装置
CN109949438A (zh) * 2019-03-20 2019-06-28 深圳市赛梅斯凯科技有限公司 异常驾驶监测模型建立方法、装置及存储介质
US20210150827A1 (en) * 2019-11-19 2021-05-20 D.S. Raider Ltd System and method for monitoring and predicting breakdowns in vehicles
CN113119981A (zh) * 2021-04-09 2021-07-16 东风汽车集团股份有限公司 车辆主动安全控制方法、系统及存储介质
CN113291311A (zh) * 2021-05-28 2021-08-24 同济大学 一种紧急避撞工况下驾驶员异常行为检测方法及存储介质
CN113901979A (zh) * 2021-08-24 2022-01-07 浙江迪嘉智能技术有限公司 一种驾驶趋势预测方法及系统
CN113815631A (zh) * 2021-09-10 2021-12-21 华人运通(江苏)技术有限公司 基于历史驾驶行为的驾驶提醒方法、装置、设备及介质
CN113963533A (zh) * 2021-09-15 2022-01-21 上海钧正网络科技有限公司 驾驶行为异常检测方法、装置、电子设备、服务器及介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117644870A (zh) * 2024-01-30 2024-03-05 吉林大学 一种基于情景感知的驾驶焦虑检测与车辆控制方法及系统
CN117644870B (zh) * 2024-01-30 2024-03-26 吉林大学 一种基于情景感知的驾驶焦虑检测与车辆控制方法及系统

Similar Documents

Publication Publication Date Title
US11941873B2 (en) Determining drivable free-space for autonomous vehicles
US11545033B2 (en) Evaluation framework for predicted trajectories in autonomous driving vehicle traffic prediction
US10981567B2 (en) Feature-based prediction
CN115206132B (zh) 基于道路使用者意图预测的潜在碰撞警告系统
CN110248861B (zh) 在车辆操纵过程中使用机器学习模型来引导车辆
US20230259792A1 (en) Object Association for Autonomous Vehicles
US20210197720A1 (en) Systems and methods for incident detection using inference models
US20200406894A1 (en) System and method for determining a target vehicle speed
US10696308B2 (en) Road condition heads up display
US11458991B2 (en) Systems and methods for optimizing trajectory planner based on human driving behaviors
CN108089571A (zh) 用于预测无人驾驶车辆的车辆交通行为以做出驾驶决策的方法和系统
CN114040869A (zh) 用于倒行车辆的规划适应
US20210191394A1 (en) Systems and methods for presenting curated autonomy-system information of a vehicle
CN108859938A (zh) 用于自动驾驶车辆的自动车辆应急灯控制的方法和系统
CN111476079A (zh) 将用于对象检测的地图特征与lidar合并的全面且有效的方法
US10909377B2 (en) Tracking objects with multiple cues
IL267885A (en) Methods and systems for calculating the presence of items
US20200283014A1 (en) Continual Planning and Metareasoning for Controlling an Autonomous Vehicle
WO2023230740A1 (zh) 一种异常驾驶行为识别的方法、装置和交通工具
US11498591B2 (en) System and method for providing adaptive trust calibration in driving automation
CN116261649A (zh) 一种车辆行驶意图预测方法、装置、终端及存储介质
CN115175841A (zh) 自主车辆的行为规划
EP3454269A1 (en) Planning autonomous motion
CN111655561A (zh) 无需地图和定位的自动驾驶车辆的拐角协商方法
US20210405641A1 (en) Detecting positioning of a sensor system associated with a 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: 22944084

Country of ref document: EP

Kind code of ref document: A1