WO2023060586A1 - 自动驾驶指令生成模型优化方法、装置、设备及存储介质 - Google Patents
自动驾驶指令生成模型优化方法、装置、设备及存储介质 Download PDFInfo
- Publication number
- WO2023060586A1 WO2023060586A1 PCT/CN2021/124192 CN2021124192W WO2023060586A1 WO 2023060586 A1 WO2023060586 A1 WO 2023060586A1 CN 2021124192 W CN2021124192 W CN 2021124192W WO 2023060586 A1 WO2023060586 A1 WO 2023060586A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- automatic driving
- driving instruction
- generation model
- instruction set
- instruction generation
- Prior art date
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000006870 function Effects 0.000 claims description 76
- 238000012549 training Methods 0.000 claims description 49
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 230000006399 behavior Effects 0.000 abstract description 15
- 238000010586 diagram Methods 0.000 description 13
- 230000007613 environmental effect Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
- B60W30/18—Propelling the vehicle
- B60W30/182—Selecting between different operative modes, e.g. comfort and performance modes
-
- 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
Definitions
- the present application relates to the field of vehicle technology, and in particular to an automatic driving command generation model optimization method, device, device and storage medium.
- the main purpose of this application is to provide an automatic driving instruction generation model optimization method, device, equipment and storage medium, aiming at solving the technical problem of how to accurately obtain the automatic driving instruction set of different driver's operation behaviors.
- the present application provides a method for optimizing an automatic driving instruction generation model, the automatic driving instruction generation model optimization method comprising:
- the manual driving instruction set includes the driving instruction input by the driver when the vehicle to be optimized is in the manual driving mode
- the automatic driving instruction set includes the vehicle to be optimized Driving instructions generated based on the automatic driving instruction generation model
- the automatic driving instruction generation model is optimized according to the manual driving instruction set and the automatic driving instruction set.
- the step of optimizing the automatic driving instruction generation model according to the manual driving instruction set and the automatic driving instruction set includes:
- the automatic driving instruction generation model is optimized according to the automatic driving instruction set and the instruction loss function.
- the step of determining the instruction loss function of the automatic driving instruction generation model according to the manual driving operating parameters and the automatic driving operating parameters includes:
- An instruction loss function of the automatic driving instruction generation model is calculated according to the manual driving operating parameters and the automatic driving operating parameters through a preset loss formula.
- the step of optimizing the automatic driving instruction generation model according to the automatic driving instruction set and the instruction loss function includes:
- the step of determining the weight update value corresponding to the automatic driving instruction generation model according to the automatic driving instruction set and the instruction loss function includes:
- a weight update value corresponding to the automatic driving instruction generation model is determined according to the historical weight value and the instruction loss function.
- the step of determining the weight update value corresponding to the automatic driving instruction generation model according to the weight history value and the instruction loss function includes:
- a weight update value corresponding to the automatic driving instruction generation model is calculated according to the weight history value and the instruction loss function through a preset weight update formula.
- the step of obtaining the manual driving instruction set and the automatic driving instruction set in the preset road section it also includes:
- a preset model initial parameter training set according to the vehicle driving path sample training subset, the manual driving operation performance parameter sample and the road environment parameter sample training subset, and the preset model initial parameter training set includes several groups Preset model initial parameters;
- the initial neural network is trained according to each set of preset model parameters and the manual driving initial instruction set corresponding to each set of preset model parameters to obtain an automatic driving instruction generation model.
- the present application also proposes an automatic driving command generation model optimization device, the automatic driving command generation model optimization device includes:
- the obtaining module is used to obtain the manual driving instruction set and the automatic driving instruction set in the preset road section, the manual driving instruction set includes the driving instruction input by the driver when the vehicle to be optimized is in the manual driving mode, and the automatic driving instruction set includes The driving instruction generated by the vehicle to be optimized based on the automatic driving instruction generation model;
- a model optimization module configured to optimize the automatic driving instruction generation model according to the manual driving instruction set and the automatic driving instruction set.
- an automatic driving instruction generation model optimization device which includes: a memory, a processor, and an automatic driving instruction generation model stored on the memory and operable on the processor
- An optimization program the automatic driving command generation model optimization program is configured to implement the steps of the automatic driving command generation model optimization method as described above.
- the present application also proposes a storage medium, on which an automatic driving command generation model optimization program is stored, and when the automatic driving command generation model optimization program is executed by a processor, the above-mentioned The steps of the automatic driving command generation model optimization method.
- the manual driving instruction set includes the driving instructions input by the driver when the vehicle to be optimized is in the manual driving mode.
- the driving instructions generated by the instruction generation model are then optimized according to the manual driving instruction set and the automatic driving instruction set to the automatic driving instruction generation model.
- the unique autopilot instruction set that has been trained is input into the vehicle control system in advance, and the vehicle can be driven directly according to the unique autopilot instruction set, resulting in the unique autopilot instruction set not being able to fit the operations of different drivers
- the automatic driving instruction generation model can be optimized according to the manual driving instruction set and the automatic driving instruction set input by the driver, so as to accurately obtain the automatic driving instruction set of different driver operation behaviors, thereby improving the driver's awareness of The experience of autonomous driving.
- FIG. 1 is a schematic structural diagram of an automatic driving instruction generation model optimization device for a hardware operating environment involved in an embodiment of the present application
- FIG. 2 is a schematic flow diagram of the first embodiment of the automatic driving instruction generation model optimization method of the present application
- FIG. 3 is a schematic diagram of building an automatic driving instruction generation model in the first embodiment of the automatic driving instruction generation model optimization method of the present application
- FIG. 4 is a schematic flow diagram of the second embodiment of the automatic driving instruction generation model optimization method of the present application.
- FIG. 5 is a schematic diagram of the automatic driving instruction generation model optimization principle diagram of the second embodiment of the automatic driving instruction generation model optimization method of the present application.
- Fig. 6 is a structural block diagram of the first embodiment of the automatic driving command generation model optimization device of the present application.
- FIG. 1 is a schematic structural diagram of an automatic driving instruction generation model optimization device for a hardware operating environment involved in an embodiment of the present application.
- the automatic driving instruction generation model optimization device may include: a processor 1001, such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005.
- the communication bus 1002 is used to realize connection and communication between these components.
- the user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface.
- the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface).
- the memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory.
- RAM Random Access Memory
- NVM Non-Volatile Memory
- the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .
- the structure shown in Figure 1 does not constitute a limitation to the automatic driving instruction generation model optimization device, and may include more or less components than those shown in the illustration, or combine certain components, or have different Part placement.
- the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and an automatic driving instruction generation model optimization program.
- the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; in the automatic driving instruction generation model optimization device of this application
- the processor 1001 and the memory 1005 can be set in the automatic driving instruction generation model optimization device, and the automatic driving instruction generation model optimization device calls the automatic driving instruction generation model optimization program stored in the memory 1005 through the processor 1001, and executes the application
- An embodiment of the present application provides a method for optimizing an automatic driving instruction generation model. Referring to FIG. 2 , FIG. 2 is a schematic flowchart of a first embodiment of the automatic driving instruction generation model optimization method of the present application.
- the automatic driving command generation model optimization method includes the following steps:
- Step S10 Obtain the manual driving instruction set and the automatic driving instruction set in the preset road section, the manual driving instruction set includes the driving instruction input by the driver when the vehicle to be optimized is in manual driving mode, the automatic driving instruction set includes the The vehicle to be optimized is based on the driving instructions generated by the automatic driving instruction generation model.
- execution subject of this embodiment may be an automatic driving instruction generation model optimization device with functions such as data processing, network communication, and program operation, or other computer devices with similar functions. This embodiment does not be restricted.
- the vehicle to be optimized can be understood as the vehicle driven by the driver, and the preset road section can be understood as the road section on which the vehicle to be optimized travels.
- the manual driving instruction set includes the driving instruction input by the driver when the vehicle to be optimized is in the manual driving mode, and the automatic driving instruction The set includes the driving instructions generated by the vehicle to be optimized based on the automatic driving instruction generation model, etc.
- the driving instruction input by the driver in the manual driving mode is the manual driving instruction set determined by the driver according to the driving operation behavior information and the vehicle's environmental variable information, in which the driving path information, the front speed Information, slope information output by the slope sensor, vehicle speed information output by the vehicle speed sensor, braking amount information output by the brake sensor, engine speed information output by the engine speed sensor, accelerator information output by the throttle sensor, and steering wheel angle output by the steering wheel angle sensor Information, etc.
- the environmental variable information of the vehicle includes the target parameters output by the camera, the point cloud data output by the lidar, the vehicle positioning data to be optimized output by the integrated inertial navigation, and the obstacle data output by the millimeter-wave radar.
- the vehicle control system can be based on the driving path information, the front vehicle speed information, the slope information output by the slope sensor, the vehicle speed information output by the vehicle speed sensor, the braking amount information output by the braking amount sensor, and the engine speed output by the engine speed sensor.
- the manual driving instruction set is determined based on the object data, and the manual driving instruction set includes an accelerator amount instruction, a steering wheel angle instruction, and a braking amount instruction, etc.
- the manual driving instruction set is the driving instruction actually output by the driver's operation behavior information and the vehicle's environmental variable information.
- the automatic driving instruction set also includes the accelerator amount instruction, the steering wheel angle instruction and the braking amount instruction.
- the initial parameter training set of the preset model includes several sets of initial parameters of the preset model, and then the corresponding manual driving is determined according to each set of initial parameters of the preset model
- the initial instruction set, and finally the initial neural network is trained according to each set of preset model parameters and the manual driving initial instruction set corresponding to each set of preset model parameters, and the automatic driving instruction generation model is obtained, so that the automatic driving instruction generation model outputs the vehicle driving The autopilot instruction set corresponding to the path.
- the manual driving operation performance parameter sample training subset includes slope A plurality of slope information output by the sensor, a plurality of vehicle speed information output by the vehicle speed sensor, a plurality of braking amount information output by the braking amount sensor, a plurality of engine speed information output by the engine speed sensor, a plurality of throttle information output by the throttle sensor, Multiple steering wheel angle information output by the steering wheel angle sensor, etc.
- the initial parameter training set of the preset model includes the speed of the preceding vehicle, vehicle driving path, slope information, vehicle speed information, braking amount information, engine speed information, accelerator information, steering wheel angle information, target object parameters, point cloud data, vehicle positioning data and Obstacle data, etc., including the speed of the front vehicle, vehicle driving path, slope information, vehicle speed information, braking amount information, engine speed information, accelerator information, steering wheel angle information, target object parameters, point cloud data, vehicle positioning data and obstacles There is a one-to-one correspondence between the data.
- the automatic driving instruction set is the driving instruction set generated by converting the initial instruction set of manual driving.
- FIG. 3 is a schematic diagram of building an automatic driving instruction generation model of the first embodiment of the automatic driving instruction generation model optimization method of the present application.
- a in FIG. The path of the vehicle and the speed of the vehicle.
- B is the manual driving operation performance parameter collection module.
- the manual driving operation performance parameter collection module can obtain slope information through the slope sensor, vehicle speed information by the vehicle speed sensor, braking amount information by the braking amount sensor, engine speed information by the engine speed sensor, throttle The sensor obtains the throttle information and the steering wheel angle sensor obtains the steering wheel angle information, etc.
- C is the road environment parameter collection module.
- the road environment parameter collection module can obtain the target object parameters through the camera, the laser radar to obtain the point cloud data, and the combined inertial navigation to obtain the vehicle positioning to be optimized.
- D is the initial neural network
- E is the automatic driving command generation module.
- the initial neural network can be trained according to the local path planning module, the manual driving operation performance parameter model, and the road environment parameter collection model to obtain the automatic driving command generation model, so that the autopilot instruction generation model outputs the corresponding autopilot instruction set.
- Step S20 Optimizing the automatic driving instruction generation model according to the manual driving instruction set and the automatic driving instruction set.
- the operating parameters of the manual driving include the actual parameters of the throttle amount, the actual parameters of the steering wheel angle, and the actual parameters of the braking amount.
- the operating parameters of the automatic driving include the predicted parameters of the throttle amount, the predicted parameters of the steering wheel angle, and the predicted parameters of the braking amount.
- the preset loss formula is:
- Loss is the instruction loss function
- t is the operating parameter of manual driving
- y is the operating parameter of automatic driving.
- the processing method of optimizing the automatic driving instruction generation model can be to judge whether the instruction loss function is greater than the preset threshold, and when the instruction loss function is greater than the preset threshold, according to the automatic driving instruction set and instruction loss
- the function determines the weight update value corresponding to the autopilot instruction generation model, and optimizes the autopilot instruction generation model according to the weight update value and the autopilot instruction set.
- the preset threshold can be customized by the user.
- the automatic driving instruction set output by the driving instruction generation model is more in line with the driver's driving operation behavior.
- the weight history value of the autopilot instruction generation model can be obtained according to the autopilot instruction set, and the weight update value corresponding to the autopilot instruction generation model can be determined according to the weight history value and the instruction loss function.
- the weight update value corresponding to the automatic driving instruction generation model is calculated through the preset weight update formula, and finally the automatic driving instruction generation model is optimized according to the weight update value and the automatic driving instruction set.
- the preset weight update formula is:
- W new is the weight update value
- W old is the weight history value
- ⁇ is the weight coefficient
- Loss is the instruction loss function
- the manual driving instruction set includes the driving instructions input by the driver when the vehicle to be optimized is in the manual driving mode.
- the automatic driving instruction generation model is optimized according to the manual driving instruction set and the automatic driving instruction set.
- the unique autopilot instruction set that has been trained is input into the vehicle control system in advance, and the vehicle can be driven directly according to the unique autopilot instruction set, resulting in the unique autopilot instruction set not being able to fit the operations of different drivers
- the automatic driving instruction generation model can be optimized according to the manual driving instruction set and the automatic driving instruction set input by the driver, so as to accurately obtain the automatic driving instruction set of different driver operation behaviors, thereby improving the driver's The experience of autonomous driving.
- FIG. 4 is a schematic flowchart of a second embodiment of an automatic driving command generation model optimization method of the present application.
- step S20 further includes:
- Step S201 Determine manual driving operating parameters according to the manual driving instruction set, and determine automatic driving operating parameters according to the automatic driving instruction set.
- the manual driving instruction set is the manual driving instruction set determined by the driver on the vehicle to be optimized according to the driving operation behavior information and the vehicle’s environmental variable information. and the actual parameters of the braking amount, the automatic driving operation parameters include the predicted parameters of the throttle amount, the predicted parameters of the steering wheel angle and the predicted parameters of the braking amount.
- Step S202 Determine an instruction loss function of the automatic driving instruction generation model according to the manual driving operating parameters and the automatic driving operating parameters.
- the first preset loss formula is:
- Step S203 Optimizing the automatic driving instruction generation model according to the automatic driving instruction set and the instruction loss function.
- the preset threshold determines whether the command loss function is greater than the preset threshold.
- the preset threshold can be customized by the user. The smaller the instruction loss function, it proves that the automatic driving instruction set output by the automatic driving instruction generation model is more suitable for the driver's driving operation behavior.
- the automatic driving instruction generation model outputting the automatic driving operation parameter b1 corresponding to the road section a is obtained through model training in advance, and when the current vehicle driver d performs manual operation on the road section a, the manual driving operation parameter is obtained c1, then calculate the instruction loss function of the automatic driving instruction generation model through the preset loss formula according to the manual driving operation parameter c1 and the automatic driving operation parameter b1.
- the automatic driving operation parameter b1 corresponding to the The driving instruction set is used as the driving instruction set of the current vehicle driver d on road section a; when the instruction loss function is greater than the preset threshold, the automatic driving instruction set and instruction loss function corresponding to the automatic driving operation parameter b1 are compared to the automatic driving instruction set on the current vehicle.
- Driving instruction generation model for optimization is used as the instruction loss function of the current vehicle driver d on road section a; when the instruction loss function is greater than the preset threshold, the automatic driving instruction set and instruction loss function corresponding to the automatic driving operation parameter b1 are compared to the automatic driving instruction set on the current vehicle.
- the automatic driving instruction generation model outputting the automatic driving operation parameter b1 corresponding to the road section a is obtained through model training in advance, and when the current vehicle driver t performs manual operation on the road section a, the manual driving operation parameter is obtained t1, then calculate the instruction loss function of the automatic driving instruction generation model through the preset loss formula according to the manual driving operating parameter t1 and the automatic driving operating parameter b1.
- the automatic driving operating parameter b1 corresponding to The driving instruction set is used as the driving instruction set of the current vehicle driver t on road section a; when the instruction loss function is greater than the preset threshold, the automatic driving instruction set and instruction loss function corresponding to the automatic driving operation parameter b1 are compared to the automatic driving instruction set on the current vehicle.
- Driving instruction generation model for optimization is used as the driving instruction set of the current vehicle driver t on road section a; when the instruction loss function is greater than the preset threshold, the automatic driving instruction set and instruction loss function corresponding to the automatic driving operation parameter b1 are compared to the automatic driving instruction set on the current vehicle.
- the weight history value of the autopilot instruction generation model can be obtained according to the autopilot instruction set, and the weight update value corresponding to the autopilot instruction generation model can be determined according to the weight history value and the instruction loss function.
- the weight update value corresponding to the automatic driving instruction generation model is calculated through the preset weight update formula, and finally the automatic driving instruction generation model is optimized according to the weight update value and the automatic driving instruction set.
- the preset weight update formula is:
- W new is the weight update value
- W old is the weight history value
- ⁇ is the weight coefficient
- Loss is the instruction loss function
- FIG. 5 is a schematic diagram of the automatic driving instruction generation model optimization schematic diagram of the second embodiment of the automatic driving instruction generation model optimization method of the present application.
- E is the automatic driving instruction generation model
- E1 is the automatic driving instruction set
- F is Manual driving instruction set
- G is the preset loss formula
- G1 is the instruction loss function
- H is the optimizer.
- the automatic driving instruction generation model outputs the automatic driving instruction set in the preset road section, the driver's manual driving instruction set in the preset road section, and then passes the preset loss according to the automatic driving instruction set and the manual driving instruction set.
- the formula calculates the instruction loss function.
- the weight history value of the automatic driving instruction generation model is obtained according to the automatic driving instruction set, and the weight update corresponding to the automatic driving instruction generation model is determined according to the weight history value and the instruction loss function. value, and then optimize the autopilot instruction generation model through the optimizer according to the weight update value and the autopilot instruction set.
- the manual driving operating parameters are first determined according to the manual driving instruction set, and the automatic driving operating parameters are determined according to the automatic driving instruction set, and then the instruction loss function of the automatic driving instruction generation model is determined according to the manual driving operating parameters and the automatic driving operating parameters , and then optimize the autopilot instruction generation model according to the autopilot instruction set and the instruction loss function.
- the trained model will not be further optimized.
- the instruction set and instruction loss function optimize the automatic driving instruction generation model, thereby improving the accuracy of the automatic driving instruction set, and then making the automatic driving instruction set output by the automatic driving instruction generation model fit the driver's driving operation behavior.
- FIG. 6 is a structural block diagram of a first embodiment of an automatic driving command generation model optimization device of the present application.
- the automatic driving instruction generation model optimization device proposed in the embodiment of the present application includes:
- the obtaining module 6001 is used to obtain the manual driving instruction set and the automatic driving instruction set in the preset road section, the manual driving instruction set includes the driving instruction input by the driver when the vehicle to be optimized is in the manual driving mode, and the automatic driving instruction set The driving instructions generated by the vehicle to be optimized based on the automatic driving instruction generation model are included.
- the vehicle to be optimized can be understood as the vehicle driven by the driver, and the preset road section can be understood as the road section on which the vehicle to be optimized travels.
- the manual driving instruction set includes the driving instruction input by the driver when the vehicle to be optimized is in the manual driving mode, and the automatic driving instruction The set includes the driving instructions generated by the vehicle to be optimized based on the automatic driving instruction generation model, etc.
- the driving instruction input by the driver in the manual driving mode is the manual driving instruction set determined by the driver according to the driving operation behavior information and the vehicle's environmental variable information, in which the driving path information, the front speed Information, slope information output by the slope sensor, vehicle speed information output by the vehicle speed sensor, braking amount information output by the brake sensor, engine speed information output by the engine speed sensor, accelerator information output by the throttle sensor, and steering wheel angle output by the steering wheel angle sensor Information, etc.
- the environmental variable information of the vehicle includes the target parameters output by the camera, the point cloud data output by the lidar, the vehicle positioning data to be optimized output by the integrated inertial navigation, and the obstacle data output by the millimeter-wave radar.
- the vehicle control system can be based on the driving path information, the front vehicle speed information, the slope information output by the slope sensor, the vehicle speed information output by the vehicle speed sensor, the braking amount information output by the braking amount sensor, and the engine speed output by the engine speed sensor.
- the manual driving instruction set is determined based on the object data, and the manual driving instruction set includes an accelerator amount instruction, a steering wheel angle instruction, and a braking amount instruction, etc.
- the manual driving instruction set is the driving instruction actually output by the driver's operation behavior information and the vehicle's environmental variable information.
- the automatic driving instruction set also includes the accelerator amount instruction, the steering wheel angle instruction and the braking amount instruction.
- the initial parameter training set of the preset model includes several sets of initial parameters of the preset model, and then the corresponding manual driving is determined according to each set of initial parameters of the preset model
- the initial instruction set, and finally the initial neural network is trained according to each set of preset model parameters and the manual driving initial instruction set corresponding to each set of preset model parameters, and the automatic driving instruction generation model is obtained, so that the automatic driving instruction generation model outputs the vehicle driving The autopilot instruction set corresponding to the path.
- the manual driving operation performance parameter sample training subset includes slope A plurality of slope information output by the sensor, a plurality of vehicle speed information output by the vehicle speed sensor, a plurality of braking amount information output by the braking amount sensor, a plurality of engine speed information output by the engine speed sensor, a plurality of throttle information output by the throttle sensor, Multiple steering wheel angle information output by the steering wheel angle sensor, etc.
- the initial parameter training set of the preset model includes the speed of the preceding vehicle, vehicle driving path, slope information, vehicle speed information, braking amount information, engine speed information, accelerator information, steering wheel angle information, target object parameters, point cloud data, vehicle positioning data and Obstacle data, etc., including the speed of the front vehicle, vehicle driving path, slope information, vehicle speed information, braking amount information, engine speed information, accelerator information, steering wheel angle information, target object parameters, point cloud data, vehicle positioning data and obstacles There is a one-to-one correspondence between the data.
- the automatic driving instruction set is the driving instruction set generated by converting the initial instruction set of manual driving.
- FIG. 3 is a schematic diagram of building an automatic driving instruction generation model of the first embodiment of the automatic driving instruction generation model optimization method of the present application.
- a in FIG. The path of the vehicle and the speed of the vehicle.
- B is the manual driving operation performance parameter collection module.
- the manual driving operation performance parameter collection module can obtain slope information through the slope sensor, vehicle speed information by the vehicle speed sensor, braking amount information by the braking amount sensor, engine speed information by the engine speed sensor, throttle The sensor obtains the throttle information and the steering wheel angle sensor obtains the steering wheel angle information, etc.
- C is the road environment parameter collection module.
- the road environment parameter collection module can obtain the target object parameters through the camera, the laser radar to obtain the point cloud data, and the combined inertial navigation to obtain the vehicle positioning to be optimized.
- D is the initial neural network
- E is the automatic driving command generation module.
- the initial neural network can be trained according to the local path planning module, the manual driving operation performance parameter model, and the road environment parameter collection model to obtain the automatic driving command generation model, so that the autopilot instruction generation model outputs the corresponding autopilot instruction set.
- a model optimization module 6002 configured to optimize the automatic driving instruction generation model according to the manual driving instruction set and the automatic driving instruction set.
- the operating parameters of the manual driving include the actual parameters of the throttle amount, the actual parameters of the steering wheel angle, and the actual parameters of the braking amount.
- the operating parameters of the automatic driving include the predicted parameters of the throttle amount, the predicted parameters of the steering wheel angle, and the predicted parameters of the braking amount.
- the preset loss formula is:
- Loss is the instruction loss function
- t is the operating parameter of manual driving
- y is the operating parameter of automatic driving.
- the processing method of optimizing the automatic driving instruction generation model can be to judge whether the instruction loss function is greater than the preset threshold, and when the instruction loss function is greater than the preset threshold, according to the automatic driving instruction set and instruction loss
- the function determines the weight update value corresponding to the autopilot instruction generation model, and optimizes the autopilot instruction generation model according to the weight update value and the autopilot instruction set.
- the preset threshold can be customized by the user.
- the automatic driving instruction set output by the driving instruction generation model is more in line with the driver's driving operation behavior.
- the weight history value of the autopilot instruction generation model can be obtained according to the autopilot instruction set, and the weight update value corresponding to the autopilot instruction generation model can be determined according to the weight history value and the instruction loss function.
- the weight update value corresponding to the automatic driving instruction generation model is calculated through the preset weight update formula, and finally the automatic driving instruction generation model is optimized according to the weight update value and the automatic driving instruction set.
- the preset weight update formula is:
- W new is the weight update value
- W old is the weight history value
- ⁇ is the weight coefficient
- Loss is the instruction loss function
- the manual driving instruction set includes the driving instructions input by the driver when the vehicle to be optimized is in the manual driving mode.
- the automatic driving instruction generation model is optimized according to the manual driving instruction set and the automatic driving instruction set.
- the unique autopilot instruction set that has been trained is input into the vehicle control system in advance, and the vehicle can be driven directly according to the unique autopilot instruction set, resulting in the unique autopilot instruction set not being able to fit the operations of different drivers
- the automatic driving instruction generation model can be optimized according to the manual driving instruction set and the automatic driving instruction set input by the driver, so as to accurately obtain the automatic driving instruction set of different driver operation behaviors, thereby improving the driver's The experience of autonomous driving.
- the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation.
- the technical solution of the present application can be embodied in the form of a software product in essence or the part that contributes to the prior art, and the computer software product is stored in a storage medium (such as read-only memory/random access memory, magnetic disk, optical disk), including several instructions to enable a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present application.
- a storage medium such as read-only memory/random access memory, magnetic disk, optical disk
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Human Computer Interaction (AREA)
- Traffic Control Systems (AREA)
Abstract
本申请公开了一种自动驾驶指令生成模型优化方法、装置、设备及存储介质，所述方法包括：获取预设路段内的人工驾驶指令集以及自动驾驶指令集，人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，自动驾驶指令集包括待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令；根据人工驾驶指令集和自动驾驶指令集对自动驾驶指令生成模型进行优化。相较于现有技术中预先将训练完成的唯一自动驾驶指令集输入至车辆控制系统内，可直接根据唯一自动驾驶指令集进行车辆行驶，而本申请中可以根据驾驶员输入的人工驾驶指令集及自动驾驶指令集对自动驾驶指令生成模型进行优化，从而精准获取不同驾驶员操作行为的自动驾驶指令集。
Description
本申请涉及车辆技术领域，尤其涉及一种自动驾驶指令生成模型优化方法、装置、设备及存储介质。
随着人工智能技术的发展，汽车电动化、智能化技术也将衍生全新的变化。按照自动驾驶等级划分，级别较高的自动驾驶功能对车辆智能化程度及功能备份有着更高的要求，现有技术中预先根据采集的道路环境信息进行训练，以获得该道路环境信息对应的唯一自动驾驶指令集，之后将唯一自动驾驶指令集输入至车辆控制系统内，驾驶员直接根据唯一自动驾驶指令集进行车辆行驶，但唯一自动驾驶指令集不能贴合不同驾驶员的操作行为，导致驾驶员对自动驾驶的体验感较差。
上述内容仅用于辅助理解本申请的技术方案，并不代表承认上述内容是现有技术。
本申请的主要目的在于提供了一种自动驾驶指令生成模型优化方法、装置、设备及存储介质，旨在解决如何精准获取不同驾驶员操作行为的自动驾驶指令集的技术问题。
为实现上述目的，本申请提供了一种自动驾驶指令生成模型优化方法，所述自动驾驶指令生成模型优化方法包括：
获取预设路段内的人工驾驶指令集以及自动驾驶指令集，所述人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，所述自动驾驶指令集包括所述待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令；
根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。在一实施例中，所述根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化的步骤，包括：
根据所述人工驾驶指令集确定人工驾驶运行参数，并根据所述自动驾驶指令集确定自动驾驶运行参数；
根据所述人工驾驶运行参数与所述自动驾驶运行参数确定所述自动驾驶指令生成模型的指令损失函数；
根据所述自动驾驶指令集和所述指令损失函数对所述自动驾驶指令生成模型进行优化。
在一实施例中，所述根据所述人工驾驶运行参数与所述自动驾驶运行参数确定所述自动驾驶指令生成模型的指令损失函数的步骤，包括：
根据所述人工驾驶运行参数与所述自动驾驶运行参数通过预设损失公式计算所述自动驾驶指令生成模型的指令损失函数。
在一实施例中，所述根据所述自动驾驶指令集和所述指令损失函数对所述自动驾驶指令生成模型进行优化的步骤，包括：
判断所述指令损失函数是否大于预设阈值；
在所述指令损失函数大于预设阈值时，根据所述自动驾驶指令集和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值；
根据所述权重更新值和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
在一实施例中，所述根据所述自动驾驶指令集和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值的步骤，包括：
根据所述自动驾驶指令集获取所述自动驾驶指令生成模型的权重历史值；
根据所述权重历史值和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值。
在一实施例中，所述根据所述权重历史值和所述指令损失函数确定所述自动驾驶指令生成 模型对应的权重更新值的步骤，包括：
根据所述权重历史值和所述指令损失函数通过预设权重更新公式计算所述自动驾驶指令生成模型对应的权重更新值。
在一实施例中，所述获取预设路段内的人工驾驶指令集以及自动驾驶指令集的步骤之前，还包括：
获取车辆行驶路径样本训练子集、人工驾驶操作性能参数样本训练子集及道路环境参数样本训练子集；
根据所述车辆行驶路径样本训练子集、所述人工驾驶操作性能参数样本及所述道路环境参数样本训练子集构建预设模型初始参数训练集，所述预设模型初始参数训练集包括若干组预设模型初始参数；
根据各组预设模型初始参数确定对应的人工驾驶初始指令集；
根据各组预设模型参数和各组预设模型参数对应的人工驾驶初始指令集对初始神经网络进行训练，获得自动驾驶指令生成模型。
为实现上述目的，本申请还提出一种自动驾驶指令生成模型优化装置，所述自动驾驶指令生成模型优化装置包括：
获取模块，用于获取预设路段内的人工驾驶指令集以及自动驾驶指令集，所述人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，所述自动驾驶指令集包括所述待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令；
模型优化模块，用于根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
为实现上述目的，本申请还提出一种自动驾驶指令生成模型优化设备，所述设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的自动驾驶指令生成模型优化程序，所述自动驾驶指令生成模型优化程序配置为实现如上文所述的自动驾驶指令生成模型优化方法的步骤。
此外，为实现上述目的，本申请还提出一种存储介质，所述存储介质上存储有自动驾驶指令生成模型优化程序，所述自动驾驶指令生成模型优化程序被处理器执行时实现如上文所述的自动驾驶指令生成模型优化方法的步骤。
本申请首先获取预设路段内的人工驾驶指令集以及自动驾驶指令集，人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，自动驾驶指令集包括待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令，之后根据人工驾驶指令集和自动驾驶指令集对自动驾驶指令生成模型进行优化。相较于现有技术中预先将训练完成的唯一自动驾驶指令集输入至车辆控制系统内，可直接根据唯一自动驾驶指令集进行车辆行驶，导致唯一自动驾驶指令集不能贴合不同驾驶员的操作行为，而本申请中可以根据驾驶员输入的人工驾驶指令集及自动驾驶指令集对自动驾驶指令生成模型进行优化，从而精准获取不同驾驶员操作行为的自动驾驶指令集，进而提高了驾驶员对自动驾驶的体验感。
图1是本申请实施例方案涉及的硬件运行环境的自动驾驶指令生成模型优化设备的结构示意图；
图2为本申请自动驾驶指令生成模型优化方法第一实施例的流程示意图；
图3为本申请自动驾驶指令生成模型优化方法第一实施例的自动驾驶指令生成模型构建原理图；
图4为本申请自动驾驶指令生成模型优化方法第二实施例的流程示意图；
图5为本申请自动驾驶指令生成模型优化方法第二实施例的自动驾驶指令生成模型优化原理图；
图6为本申请自动驾驶指令生成模型优化装置第一实施例的结构框图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
本申请的实施方式
应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
参照图1，图1为本申请实施例方案涉及的硬件运行环境的自动驾驶指令生成模型优化设备结构示意图。
如图1所示，该自动驾驶指令生成模型优化设备可以包括：处理器1001，例如中央处理器(Central Processing Unit，CPU)，通信总线1002、用户接口1003，网络接口1004，存储器1005。其中，通信总线1002用于实现这些组件之间的连接通信。用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard)，可选用户接口1003还可以包括标准的有线接口、无线接口。网络接口1004可选的可以包括标准的有线接口、无线接口(如无线保真(Wireless-Fidelity，Wi-Fi)接口)。存储器1005可以是高速的随机存取存储器(Random Access Memory，RAM)，也可以是稳定的非易失性存储器(Non-Volatile Memory，NVM)，例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。
本领域技术人员可以理解，图1中示出的结构并不构成对自动驾驶指令生成模型优化设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。如图1所示，作为一种存储介质的存储器1005中可以包括操作系统、网络通信模块、用户接口模块以及自动驾驶指令生成模型优化程序。
在图1所示的自动驾驶指令生成模型优化设备中，网络接口1004主要用于与网络服务器进行数据通信；用户接口1003主要用于与用户进行数据交互；本申请自动驾驶指令生成模型优化设备中的处理器1001、存储器1005可以设置在自动驾驶指令生成模型优化设备中，所述自动驾驶指令生成模型优化设备通过处理器1001调用存储器1005中存储的自动驾驶指令生成模型优化程序，并执行本申请实施例提供的自动驾驶指令生成模型优化方法。本申请实施例提供了一种自动驾驶指令生成模型优化方法，参照图2，图2为本申请自动驾驶指令生成模型优化方法第一实施例的流程示意图。
本实施例中，所述自动驾驶指令生成模型优化方法包括以下步骤：
步骤S10：获取预设路段内的人工驾驶指令集以及自动驾驶指令集，所述人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，所述自动驾驶指令集包括所述待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令。
易于理解的是，本实施例的执行主体可以是具有数据处理、网络通讯和程序运行等功能的自动驾驶指令生成模型优化设备，也可以为其他具有相似功能的计算机设备等，本实施例并不加以限制。
待优化车辆可以理解为驾驶员驾驶的车辆，预设路段可以理解为待优化车辆行驶的路段，其中，人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，自动驾驶指令集包括待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令等。
需要说明的是，人工驾驶模式下驾驶员输入的驾驶指令为驾驶员根据驾驶操作行为信息及车辆的环境变量信息确定的人工驾驶指令集，其中根据驾驶操作行为信息可以获取行驶路径信息、前车速信息、坡度传感器输出的坡度信息、车速传感器输出的车速信息、制动量传感器输出的制动量信息、发动机转速传感器输出的发动机转速信息、油门传感器输出的油门信息及方向盘角度传感器输出的方向盘角度信息等，车辆的环境变量信息包括摄像头输出的目标物参数、激光雷达输出的点云数据、组合惯导输出的待优化车辆定位数据及毫米波雷达输出的障碍物数据等。
在具体实现中，车辆控制系统可以根据行驶路径信息、前车速信息、坡度传感器输出的坡度信息、车速传感器输出的车速信息、制动量传感器输出的制动量信息、发动机转速传感器输出的发动机转速信息、油门传感器输出的油门信息、方向盘角度传感器输出的方向盘角度信息、摄像头输出的目标物参数、激光雷达输出的点云数据、组合惯导输出的待优化车辆定位数据集毫米波雷达输出的障碍物数据确定人工驾驶指令集，该人工驾驶指令集包括油门量指令、方向盘转角指令及制动量指令等。
应理解的是，人工驾驶指令集为驾驶员操作行为信息及车辆的环境变量信息实际输出的驾驶指令，自动驾驶指令集也包括油门量指令、方向盘转角指令及制动量指令，自动驾驶指令集为待优化车辆基于自动驾驶指令生成模型所生成的初始自动驾驶指令，根据车辆行驶路径样本训练子集、人工驾驶操作性能参数样本训练子集、道路环境参数样本训练子集及人工驾驶初始指令集对初始神经网络模型训练，以获得初始自动驾驶指令。
为了能够精准获取自动驾驶指令集，需要获取车辆行驶路径样本训练子集、人工驾驶操作性能参数样本训练子集及道路环境参数样本训练子集，然后根据车辆行驶路径样本训练子集、人工驾驶操作性能参数样本及道路环境参数样本训练子集构建预设模型初始参数训练集，预设模型初始参数训练集包括若干组预设模型初始参数，之后根据各组预设模型初始参数确定对应的人工驾驶初始指令集，最后根据各组预设模型参数和各组预设模型参数对应的人工驾驶初始指令集对初始神经网络进行训练，获得自动驾驶指令生成模型，以使自动驾驶指令生成模型输出车辆行驶路径对应的自动驾驶指令集。
车辆行驶路径样本训练子集中存在多个车辆行驶路径样本，人工驾驶操作性能参数样本训练子集中存在不同驾驶员的人工驾驶操作性能参数样本训练子集，人工驾驶操作性能参数样本训练子集包括坡度传感器输出的多个坡度信息、车速传感器输出的多个车速信息、制动量传感器输出的多个制动量信息、发动机转速传感器输出的多个发动机转速信息、油门传感器输出的多个油门信息、方向盘角度传感器输出的多个方向盘角度信息等，道路环境参数样本训练子集中存在摄像头输出的多个目标物参数、激光雷达输出的多个点云数据、组合惯导输出的多个待优化车辆定位数据及毫米波雷达输出的多个障碍物数据等。
预设模型初始参数训练集中包括前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据及障碍物数据等，其中前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据及障碍物数据存在一一对应的关系。
在本实施例中，可以根据前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据及障碍物数据生成人工驾驶初始指令集，之后根据前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据、障碍物数据及人工驾驶初始指令集对初始神经网络进行训练，获得自动驾驶指令生成模型，以使驾驶员基于行驶路径和自动驾驶指令生成模型输出自动驾驶指令集，可以理解的是，自动驾驶指令集是根据人工驾驶初始指令集进行转换所生成的驾驶指令集。
参考图3，图3为本申请自动驾驶指令生成模型优化方法第一实施例的自动驾驶指令生成模型构建原理图，图3中A为局部路径规划模块，局部路径规划模块可以获取前车速度、车辆行驶路径及本车速度。B为人工驾驶操作性能参数采集模块，人工驾驶操作性能参数采集模块可以通过坡度传感器获取坡度信息、车速传感器获取车速信息、制动量传感器获取制动量信息、发动机转速传感器获取发动机转速信息、油门传感器获取油门信息及方向盘角度传感器获取方向盘角度信息等，C为道路环境参数采集模块，道路环境参数采集模块可以通过摄像头获取目标物参数、激光雷达获取点云数据、组合惯导获取待优化车辆定位数据及毫米波雷达获取障碍物数据等。D为初始神经网络，E为自动驾驶指令生成模块，在本实施中，可以根据局部路径规划模块、人工驾驶操作性能参数模型、道路环境参数采集模型对初始神经网络进行训练，获得自动驾驶指令生成模型，以使自动驾驶指令生成模型输出对应的自动驾驶指令集。
步骤S20：根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
根据人工驾驶指令集确定人工驾驶运行参数，并根据自动驾驶指令集确定自动驾驶运行参 数，然后根据人工驾驶运行参数与自动驾驶运行参数确定自动驾驶指令生成模型的指令损失函数，最后根据自动驾驶指令集和指令损失函数对自动驾驶指令生成模型进行优化。
人工驾驶运行参数包括油门量实际参数、方向盘转角实际参数及制动量实际参数，自动驾驶运行参数包括油门量预测参数、方向盘转角预测参数及制动量预测参数。
根据人工驾驶运行参数与自动驾驶运行参数通过预设损失公式计算自动驾驶指令生成模型的指令损失函数；
预设损失公式为：
式中，Loss为指令损失函数，t为人工驾驶运行参数，y为自动驾驶运行参数。
根据自动驾驶指令集和指令损失函数对自动驾驶指令生成模型进行优化的处理方式可以为判断指令损失函数是否大于预设阈值，在指令损失函数大于预设阈值时，根据自动驾驶指令集和指令损失函数确定自动驾驶指令生成模型对应的权重更新值，根据权重更新值和自动驾驶指令集对自动驾驶指令生成模型进行优化，预设阈值可以为用户自定义设置，在指令损失函数越小时，证明自动驾驶指令生成模型输出的自动驾驶指令集越贴合驾驶员的驾驶操作行为等。
在具体实现中，可以根据自动驾驶指令集获取自动驾驶指令生成模型的权重历史值，根据权重历史值和指令损失函数确定自动驾驶指令生成模型对应的权重更新值。
根据权重历史值和指令损失函数通过预设权重更新公式计算自动驾驶指令生成模型对应的权重更新值，最后根据权重更新值和自动驾驶指令集对自动驾驶指令生成模型进行优化。预设权重更新公式为：
式中，W
new为权重更新值，W
old为权重历史值，η为权重系数，Loss为指令损失函数。
在本实施例中首先获取预设路段内的人工驾驶指令集以及自动驾驶指令集，人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，自动驾驶指令集包括待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令，之后根据人工驾驶指令集和自动驾驶指令集对自动驾驶指令生成模型进行优化。相较于现有技术中预先将训练完成的唯一自动驾驶指令集输入至车辆控制系统内，可直接根据唯一自动驾驶指令集进行车辆行驶，导致唯一自动驾驶指令集不能贴合不同驾驶员的操作行为，而本实施例中可以根据驾驶员输入的人工驾驶指令集及自动驾驶指令集对自动驾驶指令生成模型进行优化，从而精准获取不同驾驶员操作行为的自动驾驶指令集，进而提高了驾驶员对自动驾驶的体验感。
参考图4，图4为本申请自动驾驶指令生成模型优化方法第二实施例的流程示意图。
基于上述第一实施例，在本实施例中，所述步骤S20，还包括：
步骤S201：根据所述人工驾驶指令集确定人工驾驶运行参数，并根据所述自动驾驶指令集确定自动驾驶运行参数。
人工驾驶指令集为待优化车辆上驾驶员根据驾驶操作行为信息及车辆的环境变量信息确定的人工驾驶指令集，人工驾驶指令集包括当前驾驶员在当前路段上油门量实际参数、方向盘转角实际参数及制动量实际参数，自动驾驶运行参数包括油门量预测参数、方向盘转角预测参数及制动量预测参数。
步骤S202：根据所述人工驾驶运行参数与所述自动驾驶运行参数确定所述自动驾驶指令生成模型的指令损失函数。
根据油门量实际参数、方向盘转角实际参数及制动量实际参数与油门量预测参数、方向盘 转角预测参数及制动量预测参数通过第一预设损失公式计算自动驾驶指令生成模型的指令损失函数；
第一预设损失公式为：
式中，Loss为指令损失函数，t
11为油门量实际参数，t
12为方向盘转角实际参数，t
13为制动量实际参数，y
11为油门量预测参数，y
12为方向盘转角预测参数，y
13为制动量预测参数。步骤S203：根据所述自动驾驶指令集和所述指令损失函数对所述自动驾驶指令生成模型进行优化。
判断指令损失函数是否大于预设阈值，在指令损失函数大于预设阈值时，根据自动驾驶指令集和指令损失函数确定自动驾驶指令生成模型对应的权重更新值，根据权重更新值和自动驾驶指令集对自动驾驶指令生成模型进行优化，预设阈值可以为用户自定义设置，在指令损失函数越小时，证明自动驾驶指令生成模型输出的自动驾驶指令集越贴合驾驶员的驾驶操作行为等。
假设当前车辆行驶在路段a上，预先通过模型训练得到自动驾驶指令生成模型输出路段a对应的自动驾驶运行参数b1，在当前车辆驾驶员d在路段a上进行人工操作时，获取人工驾驶运行参数c1，则根据人工驾驶运行参数c1和自动驾驶运行参数b1通过预设损失公式计算自动驾驶指令生成模型的指令损失函数，在指令损失函数小于预设阈值时，将自动驾驶运行参数b1对应的自动驾驶指令集作为当前车辆驾驶员d在路段a上的驾驶指令集；在指令损失函数大于预设阈值时，将自动驾驶运行参数b1对应的自动驾驶指令集和指令损失函数对当前车辆上的自动驾驶指令生成模型进行优化。
假设当前车辆行驶在路段a上，预先通过模型训练得到自动驾驶指令生成模型输出路段a对应的自动驾驶运行参数b1，在当前车辆驾驶员t在路段a上进行人工操作时，获取人工驾驶运行参数t1，则根据人工驾驶运行参数t1和自动驾驶运行参数b1通过预设损失公式计算自动驾驶指令生成模型的指令损失函数，在指令损失函数小于预设阈值时，将自动驾驶运行参数b1对应的自动驾驶指令集作为当前车辆驾驶员t在路段a上的驾驶指令集；在指令损失函数大于预设阈值时，将自动驾驶运行参数b1对应的自动驾驶指令集和指令损失函数对当前车辆上的自动驾驶指令生成模型进行优化。
在具体实现中，可以根据自动驾驶指令集获取自动驾驶指令生成模型的权重历史值，根据权重历史值和指令损失函数确定自动驾驶指令生成模型对应的权重更新值。
根据权重历史值和指令损失函数通过预设权重更新公式计算自动驾驶指令生成模型对应的权重更新值，最后根据权重更新值和自动驾驶指令集对自动驾驶指令生成模型进行优化。预设权重更新公式为：
式中，W
new为权重更新值，W
old为权重历史值，η为权重系数，Loss为指令损失函数。
参考图5，图5为本申请自动驾驶指令生成模型优化方法第二实施例的自动驾驶指令生成模型优化原理图，图5中E为自动驾驶指令生成模型，E1为自动驾驶指令集，F为人工驾驶指令集，G为预设损失公式，G1为指令损失函数，H为优化器。在具体实现中，在预设路段中自动驾驶指令生成模型输出自动驾驶指令集，在该预设路段中驾驶员的人工驾驶指令集，之后根据自动驾驶指令集和人工驾驶指令集通过预设损失公式计算指令损失函数，在指令损失函数大于预设阈值时，根据自动驾驶指令集获取自动驾驶指令生成模型的权重历史值，根据权重历史值和指令损失函数确定自动驾驶指令生成模型对应的权重更新值， 之后根据权重更新值和自动驾驶指令集通过优化器对自动驾驶指令生成模型进行优化。
在本实施例中首先根据人工驾驶指令集确定人工驾驶运行参数，并根据自动驾驶指令集确定自动驾驶运行参数，然后根据人工驾驶运行参数与自动驾驶运行参数确定自动驾驶指令生成模型的指令损失函数，之后根据自动驾驶指令集和指令损失函数对自动驾驶指令生成模型进行优化，相较于现有技术中不会对训练好的模型进一步优化，而本实施例中根据人工驾驶指令集、自动驾驶指令集及指令损失函数对自动驾驶指令生成模型进行优化，从而提高自动驾驶指令集的精准性，进而使自动驾驶指令生成模型输出的自动驾驶指令集贴合驾驶员的驾驶操作行为。
参照图6，图6为本申请自动驾驶指令生成模型优化装置第一实施例的结构框图。
如图6所示，本申请实施例提出的自动驾驶指令生成模型优化装置包括：
获取模块6001，用于获取预设路段内的人工驾驶指令集以及自动驾驶指令集，所述人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，所述自动驾驶指令集包括所述待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令。
待优化车辆可以理解为驾驶员驾驶的车辆，预设路段可以理解为待优化车辆行驶的路段，其中，人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，自动驾驶指令集包括待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令等。
需要说明的是，人工驾驶模式下驾驶员输入的驾驶指令为驾驶员根据驾驶操作行为信息及车辆的环境变量信息确定的人工驾驶指令集，其中根据驾驶操作行为信息可以获取行驶路径信息、前车速信息、坡度传感器输出的坡度信息、车速传感器输出的车速信息、制动量传感器输出的制动量信息、发动机转速传感器输出的发动机转速信息、油门传感器输出的油门信息及方向盘角度传感器输出的方向盘角度信息等，车辆的环境变量信息包括摄像头输出的目标物参数、激光雷达输出的点云数据、组合惯导输出的待优化车辆定位数据及毫米波雷达输出的障碍物数据等。
在具体实现中，车辆控制系统可以根据行驶路径信息、前车速信息、坡度传感器输出的坡度信息、车速传感器输出的车速信息、制动量传感器输出的制动量信息、发动机转速传感器输出的发动机转速信息、油门传感器输出的油门信息、方向盘角度传感器输出的方向盘角度信息、摄像头输出的目标物参数、激光雷达输出的点云数据、组合惯导输出的待优化车辆定位数据集毫米波雷达输出的障碍物数据确定人工驾驶指令集，该人工驾驶指令集包括油门量指令、方向盘转角指令及制动量指令等。
应理解的是，人工驾驶指令集为驾驶员操作行为信息及车辆的环境变量信息实际输出的驾驶指令，自动驾驶指令集也包括油门量指令、方向盘转角指令及制动量指令，自动驾驶指令集为待优化车辆基于自动驾驶指令生成模型所生成的初始自动驾驶指令，根据车辆行驶路径样本训练子集、人工驾驶操作性能参数样本训练子集、道路环境参数样本训练子集及人工驾驶初始指令集对初始神经网络模型训练，以获得初始自动驾驶指令。
为了能够精准获取自动驾驶指令集，需要获取车辆行驶路径样本训练子集、人工驾驶操作性能参数样本训练子集及道路环境参数样本训练子集，然后根据车辆行驶路径样本训练子集、人工驾驶操作性能参数样本及道路环境参数样本训练子集构建预设模型初始参数训练集，预设模型初始参数训练集包括若干组预设模型初始参数，之后根据各组预设模型初始参数确定对应的人工驾驶初始指令集，最后根据各组预设模型参数和各组预设模型参数对应的人工驾驶初始指令集对初始神经网络进行训练，获得自动驾驶指令生成模型，以使自动驾驶指令生成模型输出车辆行驶路径对应的自动驾驶指令集。
车辆行驶路径样本训练子集中存在多个车辆行驶路径样本，人工驾驶操作性能参数样本训练子集中存在不同驾驶员的人工驾驶操作性能参数样本训练子集，人工驾驶操作性能参数样本训练子集包括坡度传感器输出的多个坡度信息、车速传感器输出的多个车速信息、制动量传感器输出的多个制动量信息、发动机转速传感器输出的多个发动机转速信息、油门传感器输出的多个油门信息、方向盘角度传感器输出的多个方向盘角度信息等，道路环境 参数样本训练子集中存在摄像头输出的多个目标物参数、激光雷达输出的多个点云数据、组合惯导输出的多个待优化车辆定位数据及毫米波雷达输出的多个障碍物数据等。
预设模型初始参数训练集中包括前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据及障碍物数据等，其中前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据及障碍物数据存在一一对应的关系。
在本实施例中，可以根据前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据及障碍物数据生成人工驾驶初始指令集，之后根据前车速度、车辆行驶路径、坡度信息、车速信息、制动量信息、发动机转速信息、油门信息、方向盘角度信息、目标物参数、点云数据、车辆定位数据、障碍物数据及人工驾驶初始指令集对初始神经网络进行训练，获得自动驾驶指令生成模型，以使驾驶员基于行驶路径和自动驾驶指令生成模型输出自动驾驶指令集，可以理解的是，自动驾驶指令集是根据人工驾驶初始指令集进行转换所生成的驾驶指令集。
参考图3，图3为本申请自动驾驶指令生成模型优化方法第一实施例的自动驾驶指令生成模型构建原理图，图3中A为局部路径规划模块，局部路径规划模块可以获取前车速度、车辆行驶路径及本车速度。B为人工驾驶操作性能参数采集模块，人工驾驶操作性能参数采集模块可以通过坡度传感器获取坡度信息、车速传感器获取车速信息、制动量传感器获取制动量信息、发动机转速传感器获取发动机转速信息、油门传感器获取油门信息及方向盘角度传感器获取方向盘角度信息等，C为道路环境参数采集模块，道路环境参数采集模块可以通过摄像头获取目标物参数、激光雷达获取点云数据、组合惯导获取待优化车辆定位数据及毫米波雷达获取障碍物数据等。D为初始神经网络，E为自动驾驶指令生成模块，在本实施中，可以根据局部路径规划模块、人工驾驶操作性能参数模型、道路环境参数采集模型对初始神经网络进行训练，获得自动驾驶指令生成模型，以使自动驾驶指令生成模型输出对应的自动驾驶指令集。
模型优化模块6002，用于根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
根据人工驾驶指令集确定人工驾驶运行参数，并根据自动驾驶指令集确定自动驾驶运行参数，然后根据人工驾驶运行参数与自动驾驶运行参数确定自动驾驶指令生成模型的指令损失函数，最后根据自动驾驶指令集和指令损失函数对自动驾驶指令生成模型进行优化。
人工驾驶运行参数包括油门量实际参数、方向盘转角实际参数及制动量实际参数，自动驾驶运行参数包括油门量预测参数、方向盘转角预测参数及制动量预测参数。
根据人工驾驶运行参数与自动驾驶运行参数通过预设损失公式计算自动驾驶指令生成模型的指令损失函数；
预设损失公式为：
式中，Loss为指令损失函数，t为人工驾驶运行参数，y为自动驾驶运行参数。
根据自动驾驶指令集和指令损失函数对自动驾驶指令生成模型进行优化的处理方式可以为判断指令损失函数是否大于预设阈值，在指令损失函数大于预设阈值时，根据自动驾驶指令集和指令损失函数确定自动驾驶指令生成模型对应的权重更新值，根据权重更新值和自动驾驶指令集对自动驾驶指令生成模型进行优化，预设阈值可以为用户自定义设置，在指令损失函数越小时，证明自动驾驶指令生成模型输出的自动驾驶指令集越贴合驾驶员的驾驶操作行为等。
在具体实现中，可以根据自动驾驶指令集获取自动驾驶指令生成模型的权重历史值，根据权重历史值和指令损失函数确定自动驾驶指令生成模型对应的权重更新值。
根据权重历史值和指令损失函数通过预设权重更新公式计算自动驾驶指令生成模型对应的权重更新值，最后根据权重更新值和自动驾驶指令集对自动驾驶指令生成模型进行优化。预设权重更新公式为：
式中，W
new为权重更新值，W
old为权重历史值，η为权重系数，Loss为指令损失函数。
在本实施例中首先获取预设路段内的人工驾驶指令集以及自动驾驶指令集，人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，自动驾驶指令集包括待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令，之后根据人工驾驶指令集和自动驾驶指令集对自动驾驶指令生成模型进行优化。相较于现有技术中预先将训练完成的唯一自动驾驶指令集输入至车辆控制系统内，可直接根据唯一自动驾驶指令集进行车辆行驶，导致唯一自动驾驶指令集不能贴合不同驾驶员的操作行为，而本实施例中可以根据驾驶员输入的人工驾驶指令集及自动驾驶指令集对自动驾驶指令生成模型进行优化，从而精准获取不同驾驶员操作行为的自动驾驶指令集，进而提高了驾驶员对自动驾驶的体验感。
本申请自动驾驶指令生成模型优化装置的其他实施例或具体实现方式可参照上述各方法实施例，此处不再赘述。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者系统不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者系统所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者系统中还存在另外的相同要素。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。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(如只读存储器/随机存取存储器、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，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的方法。
以上仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。
Claims (10)
- 一种自动驾驶指令生成模型优化方法，其特征在于，所述自动驾驶指令生成模型优化方法包括以下步骤：获取预设路段内的人工驾驶指令集以及自动驾驶指令集，所述人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，所述自动驾驶指令集包括所述待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令；根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
- 如权利要求1所述的方法，其特征在于，所述根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化的步骤，包括：根据所述人工驾驶指令集确定人工驾驶运行参数，并根据所述自动驾驶指令集确定自动驾驶运行参数；根据所述人工驾驶运行参数与所述自动驾驶运行参数确定所述自动驾驶指令生成模型的指令损失函数；根据所述自动驾驶指令集和所述指令损失函数对所述自动驾驶指令生成模型进行优化。
- 如权利要求2所述的方法，其特征在于，所述根据所述人工驾驶运行参数与所述自动驾驶运行参数确定所述自动驾驶指令生成模型的指令损失函数的步骤，包括：根据所述人工驾驶运行参数与所述自动驾驶运行参数通过预设损失公式计算所述自动驾驶指令生成模型的指令损失函数。
- 如权利要求2或3所述的方法，其特征在于，所述根据所述自动驾驶指令集和所述指令损失函数对所述自动驾驶指令生成模型进行优化的步骤，包括：判断所述指令损失函数是否大于预设阈值；在所述指令损失函数大于预设阈值时，根据所述自动驾驶指令集和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值；根据所述权重更新值和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
- 如权利要求4所述的方法，其特征在于，所述根据所述自动驾驶指令集和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值的步骤，包括：根据所述自动驾驶指令集获取所述自动驾驶指令生成模型的权重历史值；根据所述权重历史值和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值。
- 如权利要求5所述的方法，其特征在于，所述根据所述权重历史值和所述指令损失函数确定所述自动驾驶指令生成模型对应的权重更新值的步骤，包括：根据所述权重历史值和所述指令损失函数通过预设权重更新公式计算所述自动驾驶指令生成模型对应的权重更新值。
- 如权利要求1所述的方法，其特征在于，所述获取预设路段内的人工驾驶指令集以及自动驾驶指令集的步骤之前，还包括：获取车辆行驶路径样本训练子集、人工驾驶操作性能参数样本训练子集及道路环境参数样本训练子集；根据所述车辆行驶路径样本训练子集、所述人工驾驶操作性能参数样本及所述道路环境参数样本训练子集构建预设模型初始参数训练集，所述预设模型初始参数训练集包括若干组预设模型初始参数；根据各组预设模型初始参数确定对应的人工驾驶初始指令集；根据各组预设模型参数和各组预设模型参数对应的人工驾驶初始指令集对初始神经网络进行训练，获得自动驾驶指令生成模型。
- 一种自动驾驶指令生成模型优化装置，其特征在于，所述自动驾驶指令生成模型优化装置包括：获取模块，用于获取预设路段内的人工驾驶指令集以及自动驾驶指令集，所述人工驾驶指令集包括待优化车辆处于人工驾驶模式下驾驶员输入的驾驶指令，所述自动驾驶指令集包括所述待优化车辆基于自动驾驶指令生成模型所生成的驾驶指令；模型优化模块，用于根据所述人工驾驶指令集和所述自动驾驶指令集对所述自动驾驶指令生成模型进行优化。
- 一种自动驾驶指令生成模型优化设备，其特征在于，所述设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的自动驾驶指令生成模型优化程序，所述自动驾驶指令生成模型优化程序配置为实现如权利要求1至7中任一项所述的自动驾驶指令生成模型优化方法的步骤。
- 一种存储介质，其特征在于，所述存储介质上存储有自动驾驶指令生成模型优化程序，所述自动驾驶指令生成模型优化程序被处理器执行时实现如权利要求1至7任一项所述的自动驾驶指令生成模型优化方法的步骤。
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202180098578.3A CN117396389A (zh) | 2021-10-15 | 2021-10-15 | 自动驾驶指令生成模型优化方法、装置、设备及存储介质 |
| PCT/CN2021/124192 WO2023060586A1 (zh) | 2021-10-15 | 2021-10-15 | 自动驾驶指令生成模型优化方法、装置、设备及存储介质 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2021/124192 WO2023060586A1 (zh) | 2021-10-15 | 2021-10-15 | 自动驾驶指令生成模型优化方法、装置、设备及存储介质 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2023060586A1 true WO2023060586A1 (zh) | 2023-04-20 |
Family
ID=85987244
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2021/124192 WO2023060586A1 (zh) | 2021-10-15 | 2021-10-15 | 自动驾驶指令生成模型优化方法、装置、设备及存储介质 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN117396389A (zh) |
| WO (1) | WO2023060586A1 (zh) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117146832A (zh) * | 2023-10-31 | 2023-12-01 | 北京佳格天地科技有限公司 | 融合无线通信和人工智能的农机自动驾驶控制方法及系统 |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107200017A (zh) * | 2017-05-22 | 2017-09-26 | 北京联合大学 | 一种基于深度学习的无人驾驶车辆控制系统 |
| CN110562265A (zh) * | 2019-08-19 | 2019-12-13 | 中国第一汽车股份有限公司 | 一种车辆驾驶控制系统及其控制方法 |
| CN111976741A (zh) * | 2019-05-23 | 2020-11-24 | 现代自动车株式会社 | 用于控制自动驾驶车辆的装置和方法 |
-
2021
- 2021-10-15 CN CN202180098578.3A patent/CN117396389A/zh active Pending
- 2021-10-15 WO PCT/CN2021/124192 patent/WO2023060586A1/zh active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107200017A (zh) * | 2017-05-22 | 2017-09-26 | 北京联合大学 | 一种基于深度学习的无人驾驶车辆控制系统 |
| CN111976741A (zh) * | 2019-05-23 | 2020-11-24 | 现代自动车株式会社 | 用于控制自动驾驶车辆的装置和方法 |
| CN110562265A (zh) * | 2019-08-19 | 2019-12-13 | 中国第一汽车股份有限公司 | 一种车辆驾驶控制系统及其控制方法 |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117146832A (zh) * | 2023-10-31 | 2023-12-01 | 北京佳格天地科技有限公司 | 融合无线通信和人工智能的农机自动驾驶控制方法及系统 |
| CN117146832B (zh) * | 2023-10-31 | 2024-01-02 | 北京佳格天地科技有限公司 | 融合无线通信和人工智能的农机自动驾驶控制方法及系统 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117396389A (zh) | 2024-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7000638B2 (ja) | 車両制御方法、装置、コンピュータ機器及び記憶媒体 | |
| WO2021238303A1 (zh) | 运动规划的方法与装置 | |
| CN114358128B (zh) | 一种训练端到端的自动驾驶策略的方法 | |
| US20220363259A1 (en) | Method for generating lane changing decision-making model, method for lane changing decision-making of unmanned vehicle and electronic device | |
| CN113515463B (zh) | 自动化测试方法、装置、计算机设备和存储介质 | |
| CN111679660B (zh) | 一种融合类人驾驶行为的无人驾驶深度强化学习方法 | |
| WO2023082850A1 (zh) | 行人轨迹预测方法、装置及存储介质 | |
| KR20210090580A (ko) | 경로 계획 모델 생성 방법, 장치, 기기 및 저장 매체 | |
| KR20220054755A (ko) | 운전 행위 습관 결정, 차량 주행 제어 방법, 장치 및 기기 | |
| CN113415288B (zh) | 分段式纵向车速规划方法、装置、设备及存储介质 | |
| CN112965917B (zh) | 用于自动驾驶的测试方法、装置、设备和存储介质 | |
| CN114644016A (zh) | 车辆自动驾驶决策方法、装置、车载终端和存储介质 | |
| CN113119999B (zh) | 自动驾驶特征的确定方法、装置、设备、介质及程序产品 | |
| WO2023060586A1 (zh) | 自动驾驶指令生成模型优化方法、装置、设备及存储介质 | |
| CN113264064A (zh) | 用于交叉路口场景的自动驾驶方法及相关设备 | |
| CN114771525A (zh) | 一种车辆的自动换道方法及系统 | |
| EP4330107A1 (en) | Motion planning | |
| CN118246615B (zh) | 基于人工智能的智慧园区管理系统、方法及其存储介质 | |
| WO2021258847A1 (zh) | 一种驾驶决策方法、装置及芯片 | |
| CN118092525A (zh) | 一种自动驾驶车辆的速度控制方法、装置、设备及介质 | |
| CN116442787A (zh) | 一种电动汽车能耗预警方法、装置、介质及设备 | |
| CN111738046A (zh) | 对用于基于深度学习的装置的学习的虚拟世界模拟器的物理引擎进行标定的方法及装置 | |
| CN118095482A (zh) | 一种自动驾驶算法训练方法和装置 | |
| CN118579109B (zh) | 一种基于边缘计算的车辆自动驾驶轨迹规划方法及系统 | |
| WO2022052556A1 (zh) | 一种车辆行为预测方法、装置及车辆 |
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: 21960317 Country of ref document: EP Kind code of ref document: A1 |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 202180098578.3 Country of ref document: CN |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 21960317 Country of ref document: EP Kind code of ref document: A1 |