Method and system for motion control of unmanned vehicle based on electronic and electrical information architecture

Abstract

The application discloses an unmanned vehicle motion control method and system based on an electronic and electrical information architecture, relates to the technical field of automatic driving, and comprises the following steps: acquiring vehicle state information and vehicle parameters of an unmanned vehicle; establishing a kinematic model of the unmanned vehicle according to the vehicle state information; establishing an adaptive three-loop collision detection model of the unmanned vehicle according to the vehicle parameters; based on the kinematic model, the adaptive three-loop collision detection model and an electronic and electrical architecture model, adopting a multi-hop loop delay analysis method to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture, and obtaining boundary numerical values of loop delay; and based on the boundary numerical values of loop delay, adopting an improved double reinforcement learning algorithm to control and train the unmanned vehicle, and obtaining a motion control model of the unmanned vehicle. The application solves the problem of loop delay of the heterogeneous topology of the electronic and electrical architecture, and ensures the control stability and high efficiency of the unmanned vehicle in path tracking and safe obstacle avoidance.

Description

An unmanned vehicle motion control method and system based on electronic, electrical and information architecture Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to a motion control method and system for unmanned vehicles based on an electronic, electrical and information architecture. Background Technology

[0002] In the development of key technologies for autonomous driving, motion control systems play a crucial role in executing specific driving tasks. The target vehicle needs to generate control strategies based on reference paths and real-time external environmental information to achieve path tracking and obstacle avoidance. With the development of artificial intelligence, machine learning algorithms are increasingly being applied to the motion control optimization of autonomous vehicles. Q-learning, as a representative algorithm of reinforcement learning, has been repeatedly applied to motion control problems due to its relatively simple logical structure. However, Q-learning algorithms suffer from unstable training results due to overestimation of action values, such as slow or premature convergence, leading to a non-negligible error between the estimated and actual results, ultimately affecting the optimization effect of motion control strategies for autonomous vehicles. Furthermore, most existing motion control research focuses primarily on path tracking, without fully considering the possibility of sudden obstacle blockages during path tracking. Therefore, the coupling problem between path tracking control and dynamic collision avoidance control urgently needs further investigation.

[0003] Furthermore, leveraging its advantages of robust software updates, reduced wiring, and functional integration, domain-centralized electrical and electronic architecture is becoming another hot topic in the development of intelligent vehicles. Autonomous vehicles employing this architecture represent a significant branch of intelligent vehicle development. However, the increasing number of architectural functional components, various protocols, different network topologies, and inter-domain communication modes have led to unpredictable cross-domain multi-hop information delays and system uncertainties. These factors can all affect system performance and compromise the overall system stability of autonomous vehicles. In particular, the motion control problem of autonomous vehicles involves high-precision path tracking and real-time dynamic obstacle avoidance; in practical applications, significant heterogeneous loop delays are not negligible. Therefore, analyzing and resolving the motion control stability issues caused by heterogeneous topology loop delays in domain-centralized electrical and electronic architectures is essential. Summary of the Invention

[0004] The purpose of this application is to provide a motion control method and system for unmanned vehicles based on an electronic and electrical information architecture, which can solve the control stability problem caused by the loop delay of the heterogeneous topology of the electronic and electrical architecture.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] In a first aspect, this application provides a motion control method for unmanned vehicles based on an electronic, electrical, and information architecture, comprising:

[0007] Obtain vehicle status information and vehicle parameters of the autonomous vehicle.

[0008] Based on the vehicle state information of the unmanned vehicle, a kinematic model of the unmanned vehicle is established.

[0009] Based on the vehicle parameters of the unmanned vehicle, an adaptive three-ring collision detection model for the unmanned vehicle is established; the collision detection area in the adaptive three-ring collision detection model is obtained by establishing three rings with the center of the front axle, the center of the rear axle, and the center of the vehicle as their respective centers.

[0010] Based on the kinematic model, adaptive three-ring collision detection model, and electronic and electrical architecture model of the unmanned vehicle, a multi-hop loop delay analysis method is used to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture, obtaining the boundary values ​​of the loop delay. The electronic and electrical architecture model consists of a perception system, a positioning system, an intelligent connected system, a switch, an in-vehicle network, an autonomous driving DCU, a chassis DCU, and several domain controller units. The perception system is used to perceive the dynamic information of the external environment of the unmanned vehicle; the positioning system is used to receive the positioning information of the unmanned vehicle and the global scene map; the intelligent connected system is used to receive information from the entire traffic network; the autonomous driving DCU is used to execute intelligent driving functions; the chassis DCU is used for the operation control of the unmanned vehicle chassis; and the domain controller units include a drive unit, a braking unit, and a steering unit.

[0011] Based on the boundary value of the loop delay, an improved dual reinforcement learning algorithm is used to train the unmanned vehicle to obtain the motion control model of the unmanned vehicle.

[0012] Secondly, this application provides a motion control system for an unmanned vehicle based on an electronic, electrical, and information architecture, comprising:

[0013] The parameter acquisition module is used to acquire vehicle status information and vehicle parameters of the autonomous vehicle.

[0014] The kinematic model building module is used to build a kinematic model of the unmanned vehicle based on the vehicle state information of the unmanned vehicle.

[0015] The collision detection model building module is used to build an adaptive three-ring collision detection model for the unmanned vehicle based on the vehicle parameters of the unmanned vehicle. The collision detection area in the adaptive three-ring collision detection model is obtained by building three rings with the center of the front axle, the center of the rear axle, and the center of the vehicle as the respective centers.

[0016] The computation module is used to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture based on the kinematic model, adaptive three-ring collision detection model, and electronic and electrical architecture model of the unmanned vehicle, using a multi-hop loop delay analysis method to obtain the boundary values ​​of the loop delay. The electronic and electrical architecture model consists of a perception system, a positioning system, an intelligent connected system, a switch, an in-vehicle network, an autonomous driving DCU, a chassis DCU, and several domain controller units. The perception system is used to perceive the dynamic information of the external environment of the unmanned vehicle. The positioning system is used to receive the positioning information of the unmanned vehicle and the global scene map. The intelligent connected system is used to receive information from the entire traffic network. The autonomous driving DCU is used to execute intelligent driving functions. The chassis DCU is used for the operation control of the unmanned vehicle chassis. The domain controller units include a drive unit, a braking unit, and a steering unit.

[0017] The motion control model construction module is used to train the unmanned vehicle based on the boundary values ​​of the loop delay and an improved dual reinforcement learning algorithm to obtain the motion control model of the unmanned vehicle.

[0018] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0019] This application provides a motion control method and system for unmanned vehicles based on an electronic and electrical information architecture. The method involves the following steps: First, acquiring the vehicle state information and vehicle parameters of the unmanned vehicle; then, constructing a kinematic model of the unmanned vehicle using the vehicle state information; next, establishing an adaptive three-ring collision detection model of the unmanned vehicle based on the vehicle parameters; based on the kinematic model, the adaptive three-ring collision detection model, and the electronic and electrical architecture model, estimating the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture using a multi-hop loop delay analysis method, thereby determining the boundary values ​​of the loop delay; finally, based on the boundary values ​​of the loop delay, using an improved dual-reinforcement learning algorithm to train the unmanned vehicle's control, thereby obtaining the motion control model of the unmanned vehicle. This application effectively solves the loop delay problem in heterogeneous topology electronic and electrical architectures, ensuring the control stability and high efficiency of the unmanned vehicle in path tracking and obstacle avoidance. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 is a flowchart illustrating an unmanned vehicle motion control method based on an electronic and electrical information architecture according to an embodiment of this application.

[0022] Figure 2 shows a motion control model of an autonomous vehicle provided in an embodiment of this application.

[0023] Figure 3 shows an adaptive collision detection model for an autonomous vehicle provided in an embodiment of this application.

[0024] Figure 4 is a domain-centralized electronic and electrical architecture model for an unmanned vehicle provided in an embodiment of this application.

[0025] Figure 5 is a schematic diagram of multi-hop loop delay analysis of a domain-centralized electronic and electrical architecture provided in an embodiment of this application.

[0026] Figure 6 is a structural diagram of the motion control framework of an unmanned vehicle based on a domain-centralized electronic and electrical information architecture provided in an embodiment of this application.

[0027] Figure 7 is a schematic diagram of the functional modules of an unmanned vehicle motion control system based on an electronic and electrical information architecture according to an embodiment of this application. Detailed Implementation

[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Example 1

[0031] As shown in Figure 1, this embodiment provides a motion control method for unmanned vehicles based on an electronic, electrical, and information architecture, including:

[0032] Step 101: Obtain the vehicle status information and vehicle parameters of the unmanned vehicle.

[0033] Step 102: Establish the kinematic model of the unmanned vehicle based on the vehicle state information of the unmanned vehicle.

[0034] Step 103: Based on the vehicle parameters of the unmanned vehicle, establish an adaptive three-ring collision detection model for the unmanned vehicle; the collision detection area in the adaptive three-ring collision detection model is obtained by establishing three rings with the center of the front axle, the center of the rear axle, and the center of the vehicle as their respective centers.

[0035] Step 104: Based on the kinematic model, adaptive three-ring collision detection model, and electronic and electrical architecture model of the unmanned vehicle, a multi-hop loop delay analysis method is used to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture, obtaining the boundary values ​​of the loop delay. The electronic and electrical architecture model consists of a perception system, a positioning system, an intelligent connected system, a switch, an in-vehicle network, an autonomous driving DCU, a chassis DCU, and several domain controller units. The perception system is used to perceive the dynamic information of the external environment of the unmanned vehicle. The positioning system is used to receive the positioning information of the unmanned vehicle and the global scene map. The intelligent connected system is used to receive information from the entire traffic network. The autonomous driving DCU is used to execute intelligent driving functions. The chassis DCU is used for the operation control of the unmanned vehicle chassis. The domain controller units include a drive unit, a braking unit, and a steering unit.

[0036] Step 105: Based on the boundary value of the loop delay, an improved dual reinforcement learning algorithm is used to train the unmanned vehicle to obtain the motion control model of the unmanned vehicle.

[0037] Specifically, when executing step 102, the following can be done:

[0038] Based on the vehicle state information (coordinate information) of the unmanned vehicle, and based on the front and rear axle constraints and geometric equations, the kinematic model for the motion control of the unmanned vehicle can be expressed as the following state-space equation:

[0039]

[0040] As shown in Figure 2, XOY and xoy represent the inertial coordinate system and the local coordinate system, respectively. In the inertial coordinate system, the center coordinates of the front and rear axles of the vehicle are represented by (XOY and xoy). f ,Y f ) and (X r ,Y r The yaw angle of a vehicle is indicated by ( ). In the local coordinate system, v r and δ f These represent the speed at the center of the rear axle and the steering angle of the front axle, respectively. Additionally, l represents the vehicle wheelbase, and R represents the instantaneous turning radius at the center of the rear axle.

[0041] When performing step 103, the specific steps can be as follows:

[0042] In the motion control process of autonomous vehicles, collision detection needs to be carried out in real time to ensure driving safety.

[0043] This embodiment constructs an adaptive three-ring collision detection model, as shown in Figure 3. Assuming the target vehicle's speed is 0 m / s, the collision detection area is the original area shown on the right side of Figure 3. Based on the vehicle's geometric center, three rings are arranged in an array, covering the actual physical area occupied by the vehicle. Once an obstacle enters the collision detection area, a collision is determined to occur.

[0044] Assuming the target vehicle is traveling at a certain speed, the motion state information of the autonomous vehicle needs to be considered to dynamically adjust the boundary of the collision detection area. The real-time size of the detection area is determined by calculating the adaptive circle radius. The calculation formula is as follows:

[0045]

[0046] Where r is the original radius of the collision detection region, and κ is an adjustable weighting factor. ini and a brk,max These represent the initial driving speed and the maximum braking acceleration, respectively. Braking time t brk via v ini and a brk,max This can be obtained. If there are no other objects within the adaptive collision detection area, the driving state is determined to be safe.

[0047] Before performing step 104, the process also includes:

[0048] Modeling of the domain-centralized electronic and electrical architecture on autonomous vehicles, specifically including:

[0049] The domain-centralized electronic and electrical architecture adopted by autonomous vehicles is a typical network-physical system, realizing the connection and interaction between the network side and the physical side.

[0050] The network-physical system includes a sensing system, a positioning system, intelligent connected components, switches, an in-vehicle network, and multiple domain controller units (DCUs). The sensing system, positioning system, and intelligent connected components are connected to the autonomous driving DCU via Ethernet. The multiple domain controller units, represented by the autonomous driving DCU, are connected to the switches via Ethernet. The drive unit, braking unit, and steering unit are connected to the chassis DCU via CAN buses.

[0051] In this embodiment, the perception system is responsible for perceiving scene information and the vehicle's own state information; the positioning system is responsible for receiving the vehicle's positioning information; the intelligent connected components are responsible for receiving the driving state information of other vehicles in the external environment; the autonomous driving DCU is responsible for realizing various intelligent driving functions of the autonomous vehicle, including motion control functions; the switch is responsible for realizing information transmission between different DCUs; the chassis DCU is responsible for realizing the operation control of the autonomous vehicle chassis, and can convert the received control commands into signals that can be recognized by the actuator units; the drive unit is responsible for the vehicle's acceleration drive; the braking unit is responsible for the vehicle's deceleration braking; and the steering unit is responsible for the vehicle's driving and steering.

[0052] The electronic and electrical architecture in this embodiment includes multiple service nodes and links, as shown in Figure 4. In terms of networking, the architecture includes a sensing system (sensor unit), a positioning system (positioning unit), intelligent connected components (intelligent connected units), switches, an in-vehicle network, and multiple domain controller units (DCUs). In terms of physical components, the architecture includes vehicle mechanical parts, such as chassis actuators, etc.

[0053] During the operation of the electronic and electrical architecture, specific DCUs are responsible for the main functions of their respective functional areas. Specifically, the motion controller proposed in this embodiment is integrated into the autonomous driving DCU to collaboratively achieve path tracking and obstacle avoidance tasks. The switch acts as a data exchange center, facilitating interaction between different domain controllers.

[0054] In the process of achieving complete motion control for autonomous vehicles, the perception system receives external environmental information, vehicle status, and reference path, and sends them to the autonomous driving DCU via Ethernet. Then, the motion control system proposed in this embodiment is embedded in the autonomous driving DCU, solves for the optimal motion control command, and sends it to the chassis DCU using the transmission capability of the switch. The chassis DCU then converts the control commands into signals that the actuator units can recognize. Finally, the actuator units apply drive, braking, and steering control to the autonomous vehicle based on the control signals. As shown in Figure 5, the entire process consists of 5 service nodes (perception unit (sensor unit), autonomous driving DCU, switch, chassis DCU, actuator unit) and 4 communication links. Cross-domain information transmission is completed through 4 hops (transfer #1 to transfer #4).

[0055] Specifically, when executing step 104, the following can be done:

[0056] First, the service time of nodes and links is calculated, specifically including:

[0057] In the electrical and electronic architecture, the control loop consists of 5 service nodes and 4 communication links, and cross-domain information transmission requires 4 hops. Therefore, loop delay in heterogeneous topologies is unavoidable and has a non-negligible negative impact on motion control performance. To address this issue, this embodiment employs a multi-hop loop delay analysis method to estimate the cumulative loop delay of multiple nodes and links in a domain-centralized electrical and electronic architecture.

[0058] The cumulative loop delay is directly related to the service time of each component of the network (network links or network nodes). Service time is the time required for a node or link to complete its function, and it is calculated as follows:

[0059]

[0060] Where node τ node Service time includes task queuing time τ tq and task execution time τ ti , link τ link Service time includes message queuing time τ mq and message transmission time τ mt .

[0061] Then, the cumulative delay is calculated, as follows:

[0062] For the complete control loop under the electronic and electrical architecture of an autonomous vehicle, the cumulative delay τ of the loop can be derived from the service times of 5 service nodes and 4 communication links. delay :

[0063]

[0064] Where m equals 5, n equals 4, and τ node,i For the i-th node, τ link,j Let j be the j-th link.

[0065] Finally, based on the driving modes of these service nodes (i.e., time-driven mode TD or event-driven mode ED) and the triggering modes of these communication links (time-triggered mode TT or event-triggered mode ET), the theoretical boundary of the loop delay at any time k can be expressed as:

[0066]

[0067] In the formula, τ node,i,k Let τ be the delay of the i-th node at time k. link,j,k Let τ be the delay of the j-th link at time k. tq,i,kLet τ represent the task queuing time of the i-th node at time k. ti,i,k τ represents the task execution time of the i-th node at time k. mq,j,k τ represents the message queuing time at time k in the j-th link. mt,j,k This represents the message transmission time at time k on the j-th link. TD mode is time-driven mode, ED mode is event-driven mode, TT mode is time-triggered mode, and ET mode is event-triggered mode.

[0068] Based on the above formula, a schematic diagram of multi-hop loop delay analysis of motion control process using domain-centralized electronic and electrical architecture is shown in Figure 5.

[0069] Therefore, the theoretical boundary value of the loop delay at time k can be obtained:

[0070]

[0071] Among them, T s τ is the sampling time. node,i Let τ be the delay of the i-th node. link,j Let τ be the delay of the j-th link, m be the total number of nodes, n be the total number of links, and τ be the latency of the j-th link. tq For task queuing time, τ ti τ represents the task execution time. mq For message queuing time, τ mt τ is the message transmission time, k is the time, and τ is the time of transmission. tq,1,k Let τ be the task queuing time of the sensing system at time k. ti,1,k Let T be the task execution time of the sensing system at time k. s SUP represents the sampling time, and SUP denotes the supremum of the summation, τ mq,4,k τ is the message queuing time from the chassis DCU to the actuator unit in the domain controller unit. mt,4,k τ is the message transmission time from the chassis DCU to the actuator unit. tq,5,k Let τ be the task queuing time of the actuator unit at time k. ti,5,k Let k be the execution time of the actuator unit. In the training and optimization of the motion control system, the estimated electronic and electrical architecture loop delay values ​​mentioned above will be integrated into the algorithm design of the reinforcement learning system.

[0072] Before performing step 105, the process also includes:

[0073] The design of an improved dual-reinforcement (dual-Q) learning algorithm specifically includes:

[0074] Obtain two Q functions.

[0075] An improved dual reinforcement learning algorithm is obtained by training two Q functions using a cross-learning approach.

[0076] Specifically, this embodiment designs an improved dual-Q learning algorithm. This algorithm uses two Q functions to evaluate the value of the motion control strategy and adopts a cross-learning approach to train and update the two Q functions to solve the problem of overestimation of action values.

[0077] Specifically, when performing step 105, the following can be done:

[0078] As shown in Figure 6, an improved dual reinforcement learning algorithm is used to reconstruct the reward function of reinforcement learning based on the boundary value of the loop delay; the reconstructed reward function is then optimized and iterated to obtain the motion control model of the unmanned vehicle.

[0079] Specifically, based on an improved dual-reinforcement (dual-Q) learning algorithm, the estimated loop delay boundary values ​​are incorporated into the reconstruction of the reinforcement learning reward function to address the control stability problem caused by heterogeneous topology loop delays in domain-centralized electrical and electronic architectures. Through continuous optimization and training, a stable and efficient motion control system can ultimately be obtained.

[0080] In traditional reinforcement learning algorithms, the reward function can be expressed as:

[0081]

[0082] Where, k acu k saf k rap and k com These are the accuracy coefficient, safety coefficient, speed coefficient, and comfort coefficient, respectively. tra f obs f vel f dri These represent the tracking error function, the obstacle avoidance function, the speed evaluation function, and the comfortable driving function, respectively.

[0083] To eliminate interference from the electronic and electrical architecture loop delay during the execution of the upper control strategy, the time difference between state and control is determined based on the theoretical boundary value of the loop delay, and then the reward function is reconstructed:

[0084]

[0085] Where, r ori Let f represent the initial reward function. τdelay This represents the execution function that takes latency into account, u t-τdelay P represents a control action that takes time delay into account. referenceRepresents the reference path, Obcs represents obstacle information, and R adaptive This represents the adaptive radius for vehicle collision detection. To accumulate loop delay.

[0086] Therefore, the objective function and optimal function of this motion control problem can be expressed as:

[0087]

[0088] In the formula, ξ t Let Q represent the weight set of the Q matrix. The update mechanism for the Q matrix is ​​as follows:

[0089]

[0090] In the formula, α and γ represent the learning rate and discount factor, respectively. The Q matrix is ​​trained by updating the weight set ξ. t This is achieved by using two Q matrices (Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, Q1, Q2, Q3 A Q B The value of the control strategy is evaluated by maximizing the mean of the two Q functions. The optimal control action to be executed is obtained by maximizing the mean of the two Q functions, as shown in the following formula:

[0091]

[0092] In the formula, ξ t A ξ t B Let Q be the weight set of the Q matrix.

[0093] The two Q matrices are updated using different sample experience sets. The internal weights of the Q matrices are updated through cross-learning, and the update process can be represented as follows:

[0094]

[0095] Then, to improve the training efficiency of the Q-matrix in the reinforcement learning system, Prioritized Experience Replay (PER) is used to select experience samples for training. Compared with traditional random experience replay, PER training is more targeted, and the sampling probability of each experience sample can be calculated using the following formula:

[0096]

[0097] In the formula, TD(s,u) represents the temporal difference error (TD), used to evaluate the error between the target value function and the actual value function of a specific empirical sample. The selection priority of the samples is positively correlated with the TD error value. σ is used to avoid division failure, and μ represents the power of the priority coefficient. When the value function error approaches 0 infinitely, it signifies that the training of the reinforcement learning system proposed in this embodiment is approaching perfection.

[0098] Finally, as shown in Figure 6, the reference path, vehicle status, and external environment information are input into the trained motion control system, and the vehicle's acceleration and steering angle are output as control commands to the execution system through the electronic and electrical architecture.

[0099] Example 2

[0100] As shown in Figure 7, this embodiment provides a motion control system for an unmanned vehicle based on an electronic, electrical, and information architecture, including:

[0101] The parameter acquisition module 701 is used to acquire the vehicle status information and vehicle parameters of the unmanned vehicle.

[0102] The kinematic model building module 702 is used to build a kinematic model of the unmanned vehicle based on the vehicle state information of the unmanned vehicle.

[0103] The collision detection model establishment module 703 is used to establish an adaptive three-ring collision detection model for the unmanned vehicle based on the vehicle parameters of the unmanned vehicle. The collision detection area in the adaptive three-ring collision detection model is obtained by establishing three rings with the center of the front axle, the center of the rear axle, and the center of the vehicle as their respective centers.

[0104] The calculation module 704 is used to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture based on the kinematic model, adaptive three-ring collision detection model, and electronic and electrical architecture model of the unmanned vehicle, using a multi-hop loop delay analysis method to obtain the boundary values ​​of the loop delay. The electronic and electrical architecture model consists of a perception system, a positioning system, an intelligent connected system, a switch, an in-vehicle network, an autonomous driving DCU, a chassis DCU, and several domain controller units. The perception system is used to perceive the dynamic information of the external environment of the unmanned vehicle. The positioning system is used to receive the positioning information of the unmanned vehicle and the global scene map. The intelligent connected system is used to receive information from the entire traffic network. The autonomous driving DCU is used to execute intelligent driving functions. The chassis DCU is used for the operation control of the unmanned vehicle chassis. The domain controller units include a drive unit, a braking unit, and a steering unit.

[0105] The motion control model construction module 705 is used to train the unmanned vehicle based on the boundary value of the loop delay and an improved dual reinforcement learning algorithm to obtain the motion control model of the unmanned vehicle.

[0106] Specifically, the motion control model construction module 705 includes:

[0107] The reconstruction submodule is used to reconstruct the reward function of reinforcement learning based on the boundary value of the loop delay using an improved dual reinforcement learning algorithm.

[0108] The model building submodule is used to optimize and iterate the reconstructed reward function to obtain the motion control model of the unmanned vehicle.

[0109] In summary, this application has the following technical effects:

[0110] 1) To address the problem of efficient motion control for autonomous vehicles with a domain-centralized electronic and electrical architecture, the method provided in this application significantly solves the control stability problem caused by the loop delay of the heterogeneous topology of the electronic and electrical architecture, while ensuring high efficiency in path tracking and obstacle avoidance performance of autonomous vehicles.

[0111] 2) This application innovatively introduces a multi-hop loop delay analysis method, which accurately estimates the theoretical boundary value of the closed-loop delay and deeply integrates it into the controller optimization process, thereby effectively addressing and solving the challenges in control stability.

[0112] 3) This application applies the optimized double Q learning algorithm to the training process of motion control system. This effectively avoids the problem of overestimation of motion values ​​that may be encountered in the optimization stage of traditional Q learning method, thereby achieving a significant improvement in training optimization speed and a significant enhancement in training effect.

[0113] 4) The autonomous vehicle motion control method based on a domain-centralized electronic and electrical architecture proposed in this application achieves deep integration of the information communication mechanism of the electronic and electrical architecture with traditional motion control problems, providing solid support for fully autonomous motion control in driverless environments. This method demonstrates excellent performance in multiple dimensions, including tracking accuracy, obstacle avoidance safety, driving speed, and comfort.

[0114] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A motion control method for unmanned vehicles based on an electronic, electrical, and information architecture, characterized in that, The unmanned vehicle motion control method based on electronic and electrical information architecture includes: acquiring vehicle state information and vehicle parameters of the unmanned vehicle; establishing a kinematic model of the unmanned vehicle based on the vehicle state information; establishing an adaptive three-ring collision detection model of the unmanned vehicle based on the vehicle parameters; the collision detection area in the adaptive three-ring collision detection model is obtained by establishing three rings with the center of the front axle, the center of the rear axle, and the center of the vehicle as their respective centers; based on the kinematic model, the adaptive three-ring collision detection model, and the electronic and electrical architecture model of the unmanned vehicle, a multi-hop loop delay analysis method is used to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture to obtain the boundary values ​​of the loop delay. The electronic and electrical architecture model consists of a perception system, a positioning system, an intelligent connected system, a switch, an in-vehicle network, an autonomous driving DCU, a chassis DCU, and several domain controller units. The perception system is used to perceive dynamic information of the external environment of the autonomous vehicle. The positioning system is used to receive the positioning information of the autonomous vehicle and a global scene map. The intelligent connected system is used to receive information from the entire traffic network. The autonomous driving DCU is used to execute intelligent driving functions. The chassis DCU is used for the operation control of the autonomous vehicle chassis. The domain controller units include a drive unit, a braking unit, and a steering unit. Based on the boundary value of the loop delay, an improved dual reinforcement learning algorithm is used to train the autonomous vehicle's control, resulting in the motion control model of the autonomous vehicle.

2. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 1, characterized in that, The formulaic expression for the kinematic model is as follows: In the inertial coordinate system, (X) f ,Y f (X) represents the center coordinates of the vehicle's front axle. r ,Y r () represents the center coordinates of the rear axle. This represents the vehicle's yaw angle; in the local coordinate system, v r and δ f These represent the speed at the center of the rear axle and the steering angle of the front axle, respectively. L is the vehicle wheelbase, and R is the instantaneous turning radius at the center of the rear axle. This indicates the vehicle's steering angular velocity.

3. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 1, characterized in that, The formula for calculating the adaptive circle radius in the adaptive three-ring collision detection model is as follows: Where r is the original radius of the collision detection region, κ is an adjustable length weighting factor, and v ini and a brk,max t represents the initial speed and maximum braking acceleration, respectively. brk This refers to the braking time.

4. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 1, characterized in that, Based on the kinematic model, adaptive three-ring collision detection model, and electronic and electrical architecture model of the unmanned vehicle, a multi-hop loop delay analysis method is used to estimate the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture, obtaining the boundary values ​​of the loop delay. Specifically, this includes: according to the formula... Calculate the cumulative loop delay of multiple nodes and links; based on the cumulative loop delay, according to the formula... Calculate the boundary values ​​of the loop delay; where, The cumulative loop delay across multiple nodes and links. T represents the boundary value of the loop delay. s Sampling time, Let be the delay of the i-th node. Let τ be the delay of the j-th link, m be the total number of nodes, n be the total number of links, and τ be the latency of the j-th link. tq For task queuing time, τ ti τ represents the task execution time. mq For message queuing time, τ mt τ is the message transmission time, k is the time, and τ is the time of transmission. tq,1,k Let τ be the task queuing time of the sensing system at time k. ti,1,k Let T be the task execution time of the sensing system at time k. s SUP represents the sampling time, and SUP denotes the supremum of the summation, τ mq,4,k τ is the message queuing time from the chassis DCU to the actuator unit in the domain controller unit. mt,4,k τ is the message transmission time from the chassis DCU to the actuator unit. tq,5,k Let τ be the task queuing time of the actuator unit at time k. ti,5,k Let k be the task execution time of the executor unit.

5. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 1, characterized in that, The improved steps of the dual reinforcement learning algorithm are as follows: obtain two Q functions; train the two Q functions using a cross-learning method to obtain the improved dual reinforcement learning algorithm.

6. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 1, characterized in that, Based on the boundary value of the loop delay, an improved dual reinforcement learning algorithm is used to train the unmanned vehicle to obtain the motion control model of the unmanned vehicle. Specifically, the algorithm includes: using the improved dual reinforcement learning algorithm to reconstruct the reward function of reinforcement learning based on the boundary value of the loop delay; and optimizing and iterating the reconstructed reward function to obtain the motion control model of the unmanned vehicle.

7. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 6, characterized in that, The expressions for the objective value function and the optimal value function of the improved dual reinforcement learning algorithm are as follows: ; where ξ t This represents the weight set of the Q matrix. The boundary value representing the loop delay, s t Indicates the current state, u t γ represents the control action at the current moment, and γ represents the discount factor of the reinforcement learning algorithm.

8. The unmanned vehicle motion control method based on electronic and electrical information architecture according to claim 7, characterized in that, The formula for the reconstructed reward function is as follows: ; where r ori This represents the initial reward function. This represents the execution function that takes latency into account. P represents a control action that takes time delay into account. reference Represents the reference path, Obcs represents obstacle information, and R adaptive This indicates the adaptive radius for vehicle collision detection.

9. An unmanned vehicle motion control system based on an electronic-electrical-information architecture, used to implement the unmanned vehicle motion control method based on an electronic-electrical-information architecture as described in any one of claims 1-8, characterized in that, include: The system includes a parameter acquisition module for acquiring vehicle state information and parameters of the autonomous vehicle; a kinematic model building module for building a kinematic model of the autonomous vehicle based on its vehicle state information; a collision detection model building module for building an adaptive three-ring collision detection model of the autonomous vehicle based on its vehicle parameters; the collision detection area in the adaptive three-ring collision detection model is obtained by building three rings with the center of the front axle, the center of the rear axle, and the center of the vehicle as their respective centers; a calculation module for estimating the cumulative loop delay of multiple nodes and links in the electronic and electrical architecture based on the kinematic model, the adaptive three-ring collision detection model, and the electronic and electrical architecture model, using a multi-hop loop delay analysis method to obtain the boundary values ​​of the loop delay; the electronic and electrical architecture model consists of a perception system, a positioning system, an intelligent connected system, a switch, an in-vehicle network, an autonomous driving DCU, a chassis DCU, and several domain controller units; the perception system is used to perceive dynamic information of the external environment of the autonomous vehicle; the positioning system is used to receive the positioning information of the autonomous vehicle and a global scene map. The intelligent connected system is used to receive information from the entire transportation network; the autonomous driving DCU is used to execute intelligent driving functions; the chassis DCU is used for the operation control of the unmanned vehicle chassis; the domain controller unit includes a drive unit, a braking unit, and a steering unit. The motion control model construction module is used to train the unmanned vehicle based on the boundary values ​​of the loop delay and an improved dual reinforcement learning algorithm to obtain the motion control model of the unmanned vehicle.

10. The unmanned vehicle motion control system based on an electronic and electrical information architecture according to claim 9, characterized in that, The motion control model construction module specifically includes: a reconstruction submodule, used to reconstruct the reward function of reinforcement learning based on the boundary value of the loop delay using an improved dual reinforcement learning algorithm; and a model construction submodule, used to optimize and iterate the reconstructed reward function to obtain the motion control model of the unmanned vehicle.

Images