A multi-vehicle driving control method and system based on AI

Abstract

The present invention discloses a multi-vehicle driving control method and system based on AI, the method comprising: establishing a two-way communication topology structure for a fleet, collecting the status information of each vehicle in the fleet; adopting a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet; adopting a distributed control strategy to perform collaborative collision avoidance control between each vehicle; establishing an AI-based decision model, performing real-time analysis of the fleet status through a machine learning algorithm, and setting a reward function for optimizing vehicle control; performing multi-vehicle collaborative lane change control in an emergency situation, so that the vehicle travels along the planned trajectory; adopting a model predictive control strategy to track the trajectory of each vehicle. The system comprises a communication module, a linear control module, a distributed control module, a decision module, a lane change control module and a trajectory tracking module. The present invention can optimize the control strategy, enhance the multi-vehicle collaboration capability, and improve the safety and efficiency of the formation.

Description

Technical Field

[0001] The present invention relates to the field of unmanned vehicle control technology, and in particular to an AI-based multi-vehicle driving control method and system. Background Art

[0002] With the rapid development of unmanned driving technology, path planning has become one of the key technologies in multi-vehicle platoon control systems. The MRCAS system independently developed by the Shenyang Institute of the Chinese Academy of Sciences has innovatively combined distributed and centralized control methods, successfully realizing multi-functional control such as formation, maintenance, transformation and path planning of unmanned vehicles. This system uses remote computers to centrally control group robots to ensure that they can successfully complete tasks in complex environments.

[0003] The basic concept of path planning is to plan a collision-free motion trajectory from the initial position to the target position based on the environmental information perceived by the sensors of the unmanned vehicle. According to the degree of acquisition of environmental information, path planning can be divided into global path planning and local path planning. Global path planning mainly formulates the optimal path in a known environment, while local path planning dynamically adjusts the path according to real-time environmental information to effectively avoid obstacles.

[0004] Although significant progress has been made in the control and path planning of unmanned vehicle platoons, existing technologies still face the following major problems: (1) Insufficient algorithm intelligence. Most current path planning algorithms still rely on traditional methods and lack intelligent features. They cannot fully utilize the advantages of artificial intelligence. The application of intelligent algorithms can improve the fault tolerance, autonomy and search effect of the algorithms, but this also places higher requirements on computing hardware. (2) Insufficient collaboration among multiple vehicles. With the diversification of needs, it is difficult for a single vehicle to effectively complete complex tasks. It is necessary to achieve collaboration among multiple vehicles. However, the current research on path planning for multiple vehicles is still relatively weak, and there is a lack of effective collaborative control methods. (3) Poor adaptability to complex environments. Most existing path planning algorithms are targeted at general scenarios and lack designs for specific complex environments or unmanned vehicle states. Therefore, how to effectively implement path planning in complex environments remains an issue that needs to be addressed. Summary of the invention

[0005] In view of this, an embodiment of the present invention aims to provide an AI-based multi-vehicle driving control method and system, which can optimize the control strategy, enhance the multi-vehicle collaboration capability, and improve the safety and efficiency of the formation.

[0006] The embodiment of the present invention is achieved as follows:

[0007] An AI-based multi-vehicle driving control method, comprising:

[0008] A two-way communication topology is established for the fleet to collect status information of each vehicle in the fleet.

[0009] A linear superposition strategy is adopted to coordinate the speed and position of each vehicle in the convoy.

[0010] A distributed control strategy is adopted to carry out cooperative collision avoidance control between each vehicle.

[0011] Establish an AI-based decision-making model, conduct real-time analysis of fleet status through machine learning algorithms, and set reward functions for optimizing vehicle control.

[0012] In emergency situations, multi-vehicle coordinated lane changing control is carried out to ensure that vehicles travel along the planned trajectory.

[0013] The model predictive control strategy is used to track the trajectory of each vehicle.

[0014] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the establishment of a two-way communication topology structure for the fleet and the collection of status information of each vehicle in the fleet include:

[0015] Take each vehicle in the fleet as a node and construct an information neighborhood set for each node i. , the information neighborhood set Contains information about the leading vehicle, the following vehicle, and the lead vehicle.

[0016] The status information of each vehicle is collected in real time through on-board sensors, and the status information is regularly sent to the adjacent leading and following vehicles in standardized data packets through wireless communication technology.

[0017] Data fusion technology is used to integrate the status information of each vehicle in the fleet to obtain the global fleet status information.

[0018] Its technical effect is: by treating each vehicle in the fleet as a node and using on-board sensors to collect the status information of each vehicle in real time, real-time perception of the status of the entire fleet is achieved, and a two-way communication topology is established, so that the front vehicle, the rear vehicle and the pilot vehicle can exchange status information with each other, enhancing the information sharing ability among fleet members, facilitating more accurate speed and position coordination control, as well as collaborative collision avoidance control, and improving the system's ability to adapt to environmental changes.

[0019] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the use of a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet includes:

[0020] Establish a nonlinear longitudinal dynamic model of the vehicle to describe the relationship between acceleration and various forces ,in, is the speed of the i-th vehicle at time t, is the input acceleration applied to the vehicle, is the actual driving force acting on the wheel, is the mass of the i-th vehicle, is the headwind drag coefficient, is the rolling friction resistance, is the acceleration due to gravity.

[0021] Set the control target so that the speed of the main vehicle and the pilot vehicle are consistent ,in, The speed of the pilot car, the main car and the front car maintain a safe distance ,in, is the position of the i-1th vehicle, is the position of the i-th vehicle, is the safe distance between the i-1th vehicle and the i-th vehicle.

[0022] According to the cooperative adaptive cruise control strategy, the desired input acceleration of the i-th vehicle is controlled ,in, is the actual displacement difference between the vehicle in front and the vehicle in front, is the actual displacement difference between the vehicle and the following vehicle, is the ideal following distance between the i-th vehicle and the preceding vehicle, d is the minimum displacement difference in a stationary state, m is the following coefficient, is the distance tracking coefficient, is the speed tracking coefficient, is the speed tracking coefficient relative to the pilot vehicle.

[0023] According to the cooperative adaptive cruise control strategy, the distance error between the i-th vehicle and the front vehicle is controlled ,in, The ideal safety distance is set.

[0024] Its technical effect is: by establishing a longitudinal nonlinear dynamic model of the vehicle, the vehicle state is closely combined with the control target, and the safety distance between the team members is ensured by controlling the speed consistency of the main vehicle and the pilot vehicle, and maintaining a safe distance between the main vehicle and the front vehicle. Through collaborative adaptive cruise control, the distance error between the front vehicle and the rear vehicle, as well as the set ideal safety distance, is taken into account to enhance the coordination ability between vehicles. The vehicle can not only make corresponding adjustments according to the state of the front vehicle, but also optimize the distance relationship with the rear vehicle in real time, thereby improving the coordination of the entire team.

[0025] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the use of a distributed control strategy to perform cooperative collision avoidance control between each vehicle includes:

[0026] Establish a collision time interval indicator to describe the time required for the main vehicle to collide with the front vehicle at the current speed ,in, is the current position of the preceding vehicle, is the current vehicle position, is the length of the current vehicle, is the speed of the front vehicle, is the current vehicle speed.

[0027] Establish a front vehicle time distance index to describe the time required for the main vehicle to collide with the front vehicle after the front vehicle suddenly brakes .

[0028] According to the collision time interval index and the preceding vehicle time distance index, a collision risk index is obtained. ,in, and is the weight factor.

[0029] A collision risk threshold and a collision risk upper limit are set, wherein the collision risk threshold is lower than the collision risk upper limit.

[0030] When the collision risk indicator When the speed is lower than the collision risk threshold, the low-risk area control strategy is adopted to maintain the current speed and continue driving.

[0031] When the collision risk indicator When the speed is higher than the collision risk threshold and lower than the collision risk upper limit, the medium risk area control strategy is adopted to adjust the vehicle acceleration, reduce the speed difference, and reduce the collision risk index. The value of .

[0032] When the collision risk indicator When the risk exceeds the upper limit of collision risk, the high-risk area emergency collision avoidance strategy is adopted, and the maximum braking force is applied for braking.

[0033] The technical effect is: by establishing the collision time interval index and the front vehicle time distance index, the collision risk between the current vehicle and the front vehicle is evaluated in real time. Each vehicle makes local decisions based on its own status and the status of surrounding vehicles, which enhances the cooperation ability between vehicles. When the collision risk index exceeds the threshold, an emergency collision avoidance strategy is quickly adopted, and the maximum braking force is applied for braking.

[0034] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the high-risk area emergency collision avoidance strategy includes:

[0035] When the front vehicle brakes urgently, minimize the relative kinetic energy of the convoy, perform front-end collision avoidance, and calculate the total kinetic energy of the convoy , using the objective function Total kinetic energy of the queue The optimization constraints are , , where j is all other vehicles compared with the i-th vehicle, is the mass of the i-th vehicle, is the speed of the i-th vehicle, is the speed of the jth vehicle, is the acceleration of the i-th vehicle, is the maximum braking deceleration of the i-th vehicle, is the maximum speed of the i-th vehicle.

[0036] When the rear vehicle approaches sharply, the main vehicle accelerates in coordination to avoid rear-end collision. When the distance between the rear vehicle and the main vehicle is less than the minimum warning distance, the main vehicle brakes at the maximum deceleration. When the leading vehicle is cruising at a constant speed, if a collision is still unavoidable, the main vehicle needs to accelerate in coordination, and the additional acceleration is ,in, is the expected acceleration of the following vehicle, The maximum deceleration allowed for the following vehicle.

[0037] Its technical effect is: by calculating the total kinetic energy of the fleet and optimizing it using the objective function, the relative kinetic energy between vehicles is effectively minimized, the stability of the vehicle in emergency or sudden situations is enhanced, the severity of the collision is reduced, and safety is significantly improved. The dynamic response capability enables the vehicle to not only brake in time when facing a potential collision, but also provide braking space for the following vehicle through acceleration strategies, forming an effective coordinated response mechanism, making the overall behavior of the vehicle more coordinated, and avoiding chain reactions caused by misjudgment of a single vehicle.

[0038] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the establishment of an AI-based decision model, real-time analysis of the fleet status through a machine learning algorithm, and setting a reward function for optimizing vehicle control include:

[0039] A decision model based on machine learning is established, wherein the input of the decision model is the current state of the fleet, and the output is the corresponding probability distribution.

[0040] For the current state of the vehicle fleet, a corresponding state space vector and action space vector are designed, wherein the state space vector describes the state of the vehicle and its surrounding environment, and the action space vector describes the action performed by the vehicle.

[0041] A reward function for evaluating the effect of each of the action space vectors is set, wherein the reward function includes a safety reward, a comfort reward, an efficiency reward, a cooperation reward, and an environmental impact reward.

[0042] For the decision model, the selected action space vector is applied to the vehicle, the vehicle state is updated, and the reward obtained is calculated in real time through a reward function based on the updated state and action of the vehicle at each time step, and the calculated reward is used to update the weight of the decision model.

[0043] Its technical effect is: by constructing an action space vector, the decision-making model can choose from a variety of feasible driving actions, such as acceleration, deceleration, lane change, etc. The design of the reward function includes multiple evaluation indicators such as safety, comfort, efficiency, collaboration and environmental impact. The mechanism based on reinforcement learning enables the vehicle's behavior to continuously adapt to new driving conditions and environmental changes through continuous training and feedback cycles, achieving comprehensive improvements in state perception, intelligent decision-making, reinforcement learning, collaboration mechanism and flexibility.

[0044] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the safety reward includes a safety distance reward. and collision risk penalty ,in, is the distance between vehicle i and vehicle j, is the minimum safe distance, and is the weight factor, is the collision risk factor, is the penalty weight.

[0045] The comfort bonus includes an acceleration smoothness bonus and braking and acceleration comfort bonus ,in, is the actual acceleration of the vehicle, is the ideal acceleration, is the stability weight, is the maximum permissible braking deceleration.

[0046] The efficiency bonus includes a speed maintenance bonus and energy efficiency penalty ,in, For the ideal speed, is the efficiency weight, is the energy consumption of vehicle i, is the energy consumption penalty weight.

[0047] The collaboration reward includes information sharing reward and collaborative behavior rewards ,in, is the information sharing degree between vehicle i and vehicle j, is the information sharing weight, is the weight of collaboration capability.

[0048] The environmental impact incentives include emission control penalties ,in, is the emission of vehicle i, is the emission penalty weight.

[0049] Its technical effects are: introducing safe distance rewards and collision risk penalties, so that the vehicle can evaluate the relative distance and collision risk with other vehicles in real time during the decision-making process; by real-time monitoring of the distance between the vehicle and other vehicles, and dynamic calculation of the collision risk coefficient, the vehicle can quickly respond to environmental changes and adjust the driving strategy in real time; by setting safe distance rewards, the vehicle will consciously maintain a safe distance from other vehicles during driving, promote the overall safety and traffic efficiency of the fleet, and strengthen the collaborative work between vehicles.

[0050] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the multi-vehicle coordinated lane change control in an emergency situation so that the vehicles drive along the planned trajectory includes:

[0051] The environmental information of the current fleet is obtained through sensors. If a vehicle fails to reach the expected speed or there is an obstacle ahead, a lane change intention is generated.

[0052] Assess the status of vehicles in adjacent lanes and determine the feasibility of lane changing.

[0053] If it is determined that lane changing is feasible, trajectory planning is performed to generate a lane changing trajectory so that the vehicle can travel along the planned trajectory.

[0054] Its technical effects are: by obtaining real-time environmental information of the fleet and responding quickly to emergencies, the vehicle can effectively avoid risks in complex and dynamic environments; through coordinated lane change control, the vehicle evaluates the status of adjacent lanes in real time, determines the feasibility of lane changes, and selects the optimal driving path to avoid speed reduction and traffic congestion caused by lane changes.

[0055] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the use of a model predictive control strategy to track the trajectory of each vehicle includes:

[0056] In the fleet, a dynamic model is established to predict the state of each vehicle

[0057] ;

[0058] ,

[0059] in, is the state variable of the i-th vehicle, ( ) is the position of the vehicle, For speed, is the heading angle, T is the control period, is the control input of the i-th vehicle, is the output variable of the i-th vehicle, is the output function of the vehicle.

[0060] In each control cycle T, the dynamic model is used to track the trajectory of each vehicle, and at each time k, the time step of the specified time range is predicted. The vehicle state trajectory within is .

[0061] Construct an objective function for minimizing the error between the vehicle state trajectory and the desired trajectory ,in, is the actual output at time k+i, is the expected output at time k+i, Q is the weighting matrix used to control the output error weight, is the amount of change used to control the input, and R is the weighting matrix used to control the weight of the input change.

[0062] In each control cycle T, the value of the objective function is minimized to solve the optimization problem. ,in, is the control input vector, is the minimum value limit of the control input, The maximum value limit for the control input.

[0063] Use the first value of the optimized control sequence Update the kinetic model .

[0064] Measure the actual status data of the vehicle , calculate the error between the predicted state and the measured state .

[0065] Use the predicted state of the next step of the state error to correct it and get the corrected predicted state vector .

[0066] Its technical effects are: by establishing a dynamic model for each vehicle, predicting the state of the vehicle in the future time step, foreseeing and responding to potential changes in a complex traffic environment, solving optimization problems according to the current state and expected trajectory in each control cycle, determining the optimal control sequence, and updating the adaptability and accuracy of the model through a real-time feedback correction mechanism.

[0067] An AI-based multi-vehicle driving control system, comprising:

[0068] The communication module is used to establish a two-way communication topology for the fleet and collect status information of each vehicle in the fleet.

[0069] The linear control module is used to adopt a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet.

[0070] The distributed control module is used to adopt a distributed control strategy to perform collaborative collision avoidance control between each vehicle.

[0071] The decision-making module is used to establish an AI-based decision-making model, conduct real-time analysis of the fleet status through machine learning algorithms, and set reward functions for optimizing vehicle control.

[0072] The lane change control module is used to perform multi-vehicle coordinated lane change control in emergency situations.

[0073] The trajectory tracking module is used to track the trajectory of each vehicle using a model predictive control strategy.

[0074] The beneficial effects of the present invention are:

[0075] By establishing a two-way communication topology, the present invention enables each vehicle in the fleet to collect and share status information in real time, effectively improving the coordination between vehicles, reducing the reaction delay caused by information lag, and ensuring rapid response in emergencies. Compared with traditional single-vehicle control methods, real-time information sharing greatly enhances the adaptability and flexibility of the system.

[0076] The present invention adopts a linear superposition strategy to coordinate the speed and position control, realizes the synchronization of the relative motion state of each vehicle in the convoy, effectively maintains a safe distance and improves driving efficiency, and reduces the risk of rear-end collisions caused by improper following. Compared with the existing technology, this strategy can more effectively cope with dynamic changes during high-speed driving and ensure vehicle safety and comfort.

[0077] Through the distributed control strategy, the present invention can achieve collaborative collision avoidance control between vehicles, establish collision risk indicators and take corresponding measures according to the risk level, so as to detect potential collision risks earlier and make adjustments quickly through collective decision-making. Compared with the traditional centralized control scheme, the distributed strategy can more effectively respond to emergencies and reduce the probability of collision.

[0078] The present invention uses a decision model based on a machine learning algorithm to analyze the fleet status in real time, set a reward function to optimize vehicle control, continuously learn and self-optimize, and adapt to different traffic environments and driving habits. Compared with traditional rule-based control systems, AI models can flexibly handle changing environmental factors and improve the scientificity and accuracy of decision-making.

[0079] The multi-vehicle coordinated lane-changing control implemented in the present invention in emergency situations can quickly generate lane-changing intentions and evaluate feasibility. At the same time, a model predictive control strategy is used to track the trajectory of each vehicle, effectively responding to emergencies such as traffic jams and obstacles, improving the vehicle's driving ability and safety in complex environments, achieving accurate modeling and prediction of vehicle dynamics, and optimizing control inputs in real time, thereby improving the vehicle's driving stability and accuracy under complex road conditions and reducing safety hazards caused by trajectory deviations. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0081] Figure 1 The figure is a flow chart of the AI-based multi-vehicle driving control method of the present invention. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0083] Please refer to Figure 1 The first embodiment of the present invention provides an AI-based multi-vehicle driving control method, which includes: establishing a two-way communication topology structure for a fleet, collecting the status information of each vehicle in the fleet; adopting a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet; adopting a distributed control strategy to perform collaborative collision avoidance control between each vehicle; establishing an AI-based decision model, performing real-time analysis of the fleet status through a machine learning algorithm, and setting a reward function for optimizing vehicle control; performing multi-vehicle collaborative lane change control in emergency situations to make the vehicles travel along the planned trajectory; and adopting a model predictive control strategy to track the trajectory of each vehicle.

[0084] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the establishment of a two-way communication topology structure for the fleet and the collection of status information of each vehicle in the fleet include: taking each vehicle in the fleet as a node, constructing an information neighborhood set for each node i , the information neighborhood set Contains information about the front car, the rear car, and the pilot car; for the first car, , for the middle vehicle, , , for the last car, Through on-board sensors (including GPS, radar, lidar, etc.), the status information of each vehicle (including vehicle position, speed, acceleration, braking status, and distance from the front and rear vehicles, etc.) is collected in real time, and the status information is regularly sent to the adjacent front and rear vehicles in standardized data packets through wireless communication technology; using data fusion technology, the status information of each vehicle in the fleet is integrated to obtain the global fleet status information.

[0085] Its technical effect is: by treating each vehicle in the fleet as a node and using on-board sensors to collect the status information of each vehicle in real time, real-time perception of the status of the entire fleet is achieved, and a two-way communication topology is established, so that the front vehicle, the rear vehicle and the pilot vehicle can exchange status information with each other, enhancing the information sharing ability among fleet members, facilitating more accurate speed and position coordination control, as well as collaborative collision avoidance control, and improving the system's ability to adapt to environmental changes.

[0086] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the use of a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet includes: establishing a vehicle longitudinal nonlinear dynamic model for describing the relationship between acceleration and various forces ,in, is the speed of the i-th vehicle at time t, is the input acceleration applied to the vehicle, is the actual driving force acting on the wheel, is the mass of the i-th vehicle, is the headwind drag coefficient, is the rolling friction resistance, is the acceleration due to gravity; set the control target so that the speed of the main vehicle and the pilot vehicle remain consistent ,in, The speed of the pilot car, the main car and the front car maintain a safe distance ,in, is the position of the i-1th vehicle, is the position of the i-th vehicle, is the safe distance between the i-1th vehicle and the i-th vehicle; according to the cooperative adaptive cruise control strategy, the expected input acceleration of the i-th vehicle is controlled ,in, is the actual displacement difference between the vehicle in front and the vehicle in front, is the actual displacement difference between the vehicle and the following vehicle, is the ideal following distance between the i-th vehicle and the preceding vehicle, d is the minimum displacement difference in a stationary state, m is the following coefficient, is the distance tracking coefficient, is the speed tracking coefficient, is the speed tracking coefficient relative to the pilot vehicle; according to the cooperative adaptive cruise control strategy, the distance error between the i-th vehicle and the preceding vehicle is controlled ,in, The ideal safety distance is set.

[0087] Its technical effect is: by establishing a longitudinal nonlinear dynamic model of the vehicle, the vehicle state is closely combined with the control target, and the safety distance between the team members is ensured by controlling the speed consistency of the main vehicle and the pilot vehicle, and maintaining a safe distance between the main vehicle and the front vehicle. Through collaborative adaptive cruise control, the distance error between the front vehicle and the rear vehicle, as well as the set ideal safety distance, is taken into account to enhance the coordination ability between vehicles. The vehicle can not only make corresponding adjustments according to the state of the front vehicle, but also optimize the distance relationship with the rear vehicle in real time, thereby improving the coordination of the entire team.

[0088] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the use of a distributed control strategy to perform cooperative collision avoidance control between each vehicle includes: establishing a collision time interval index for describing the time required for the main vehicle to collide with the front vehicle at the current speed ,in, is the current position of the preceding vehicle, is the current vehicle position, is the length of the current vehicle, is the speed of the front vehicle, is the current vehicle speed; establishes a front vehicle time distance index to describe the time required for the main vehicle to collide with the front vehicle after the front vehicle suddenly brakes ; According to the collision time interval index and the front vehicle time distance index, obtain the collision risk index ,in, and is a weight factor; setting a collision risk threshold and a collision risk upper limit, wherein the collision risk threshold is lower than the collision risk upper limit; when the collision risk index When the speed is lower than the collision risk threshold, the low-risk zone control strategy is adopted to maintain the current speed and continue driving; when the collision risk index is When the speed is higher than the collision risk threshold and lower than the collision risk upper limit, the medium risk area control strategy is adopted to adjust the vehicle acceleration, reduce the speed difference, and reduce the collision risk index. value; when the collision risk index When the risk exceeds the upper limit of collision risk, the high-risk area emergency collision avoidance strategy is adopted, and the maximum braking force is applied for braking.

[0089] The technical effect is: by establishing the collision time interval index and the front vehicle time distance index, the collision risk between the current vehicle and the front vehicle is evaluated in real time. Each vehicle makes local decisions based on its own status and the status of surrounding vehicles, which enhances the cooperation ability between vehicles. When the collision risk index exceeds the threshold, an emergency collision avoidance strategy is quickly adopted, and the maximum braking force is applied for braking.

[0090] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the high-risk area emergency collision avoidance strategy includes: when the front vehicle brakes urgently, minimizing the relative kinetic energy of the convoy, performing front-end collision avoidance, and calculating the total kinetic energy of the convoy. , using the objective function Total kinetic energy of the queue The optimization constraints are , , where j is all other vehicles compared with the i-th vehicle, is the mass of the i-th vehicle, is the speed of the i-th vehicle, is the speed of the jth vehicle, is the acceleration of the i-th vehicle, is the maximum braking deceleration of the i-th vehicle, is the maximum speed of the i-th vehicle; when the rear vehicle approaches sharply, the main vehicle accelerates in coordination to avoid rear-end collision. When the distance between the rear vehicle and the main vehicle is less than the minimum warning distance, the main vehicle brakes at the maximum deceleration. When the front vehicle is cruising at a constant speed, if a collision is still unavoidable, the main vehicle needs to accelerate in coordination, and the additional acceleration is ,in, is the expected acceleration of the following vehicle, The main vehicle dynamically adjusts its own acceleration to ensure that the rear vehicle has enough space to brake, thereby reducing the risk of rear-end collision.

[0091] Its technical effect is: by calculating the total kinetic energy of the fleet and optimizing it using the objective function, the relative kinetic energy between vehicles is effectively minimized, the stability of the vehicle in emergency or sudden situations is enhanced, the severity of the collision is reduced, and safety is significantly improved. The dynamic response capability enables the vehicle to not only brake in time when facing a potential collision, but also provide braking space for the following vehicle through acceleration strategies, forming an effective coordinated response mechanism, making the overall behavior of the vehicle more coordinated, and avoiding chain reactions caused by misjudgment of a single vehicle.

[0092] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the establishment of an AI-based decision model, real-time analysis of the fleet status through a machine learning algorithm, and setting a reward function for optimizing vehicle control include: establishing a decision model based on machine learning (such as linear regression, decision tree, random forest, support vector machine or neural network), the input of the decision model is the current state of the fleet, and the output is the corresponding probability distribution; for the current state of the fleet, designing a corresponding state space vector and action space vector, the state space vector describes the state of the vehicle and its surrounding environment (such as the current speed, acceleration, position of the main vehicle, the speed, position and acceleration of the front and rear vehicles , vehicle spacing and queue length, etc.), the action space vector describes the action performed by the vehicle (such as acceleration, deceleration, maintaining speed, changing lanes, emergency braking, etc.); a reward function is set for evaluating the effect of each action space vector, and the reward function includes safety reward, comfort reward, efficiency reward, collaboration reward and environmental impact reward; for the decision model, the selected action space vector is applied to the vehicle, the vehicle state is updated, and the reward is obtained by real-time calculation of the reward function based on the updated state and action of the vehicle at each time step, and the calculated reward is used to update the weight of the decision model for selecting a better action under similar states, and the model performance is continuously optimized through repeated training and feedback cycles.

[0093] Its technical effect is: by constructing an action space vector, the decision-making model can choose from a variety of feasible driving actions, such as acceleration, deceleration, lane change, etc. The design of the reward function includes multiple evaluation indicators such as safety, comfort, efficiency, collaboration and environmental impact. The mechanism based on reinforcement learning enables the vehicle's behavior to continuously adapt to new driving conditions and environmental changes through continuous training and feedback cycles, achieving comprehensive improvements in state perception, intelligent decision-making, reinforcement learning, collaboration mechanism and flexibility.

[0094] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the safety reward includes a safety distance reward. and collision risk penalty ,in, is the distance between vehicle i and vehicle j, is the minimum safe distance, and is the weight factor, is the collision risk factor, is the penalty weight; the comfort reward includes the acceleration smoothness reward and braking and acceleration comfort bonus ,in, is the actual acceleration of the vehicle, is the ideal acceleration, is the stability weight, is the maximum allowed braking deceleration; the efficiency bonus includes a speed maintenance bonus and energy efficiency penalty ,in, For the ideal speed, is the efficiency weight, is the energy consumption of vehicle i, is the energy consumption penalty weight; the collaboration reward includes information sharing reward and collaborative behavior rewards ,in, is the information sharing degree between vehicle i and vehicle j, is the information sharing weight, is the weight of collaborative capacity; the environmental impact reward includes emission control penalty ,in, is the emission of vehicle i, is the emission penalty weight.

[0095] Its technical effects are: introducing safe distance rewards and collision risk penalties, so that the vehicle can evaluate the relative distance and collision risk with other vehicles in real time during the decision-making process; by real-time monitoring of the distance between the vehicle and other vehicles, and dynamic calculation of the collision risk coefficient, the vehicle can quickly respond to environmental changes and adjust the driving strategy in real time; by setting safe distance rewards, the vehicle will consciously maintain a safe distance from other vehicles during driving, promote the overall safety and traffic efficiency of the fleet, and strengthen the collaborative work between vehicles.

[0096] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the multi-vehicle coordinated lane changing control in an emergency situation so that the vehicles can travel along the planned trajectory includes: obtaining environmental information of the current fleet through sensors, and generating a lane changing intention if a vehicle fails to reach the expected speed or there is an obstacle ahead; evaluating the vehicle status of adjacent lanes to determine the feasibility of lane changing; if it is determined that lane changing is feasible, trajectory planning is performed to generate a lane changing trajectory so that the vehicle can travel along the planned trajectory.

[0097] Its technical effects are: by obtaining real-time environmental information of the fleet and responding quickly to emergencies, the vehicle can effectively avoid risks in complex and dynamic environments; through coordinated lane change control, the vehicle evaluates the status of adjacent lanes in real time, determines the feasibility of lane changes, and selects the optimal driving path to avoid speed reduction and traffic congestion caused by lane changes.

[0098] In a preferred embodiment of the present invention, in the above-mentioned AI-based multi-vehicle driving control method, the use of a model predictive control strategy to track the trajectory of each vehicle includes: establishing a dynamic model to predict the state of each vehicle in the fleet. , ,in, is the state variable of the i-th vehicle (including position, heading angle and speed), ( ) is the position of the vehicle, For speed, is the heading angle, T is the control period, is the control input of the i-th vehicle (including the front wheel steering angle), is the output variable of the i-th vehicle (including lateral position and heading angle), is the output function of the vehicle; in each control cycle T, the trajectory of each vehicle is tracked using the dynamic model, and at each time k, the time step of the specified time range is predicted The vehicle state trajectory within is ; Construct an objective function for minimizing the error between the vehicle state trajectory and the desired trajectory ,in, is the actual output at time k+i, is the expected output at time k+i, Q is the weighting matrix used to control the output error weight, is the change amount of the control input, R is the weighting matrix for the weight of the control input change; in each control cycle T, the value of the objective function is minimized to solve the optimization problem, ,in, is the control input vector, is the minimum value limit of the control input, The maximum value of the control input is limited; the first value of the control sequence obtained by optimization is used Update the kinetic model ;Measure the actual status data of the vehicle , calculate the error between the predicted state and the measured state ; Use the predicted state of the next step of the state error to correct and obtain the corrected predicted state vector .

[0099] Its technical effects are: by establishing a dynamic model for each vehicle, predicting the state of the vehicle in the future time step, foreseeing and responding to potential changes in a complex traffic environment, solving optimization problems according to the current state and expected trajectory in each control cycle, determining the optimal control sequence, and updating the adaptability and accuracy of the model through a real-time feedback correction mechanism.

[0100] The second embodiment of the present invention provides an AI-based multi-vehicle driving control system, which includes: a communication module, which is used to establish a two-way communication topology for the fleet and collect status information of each vehicle in the fleet; a linear control module, which is used to adopt a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet; a distributed control module, which is used to adopt a distributed control strategy to perform collaborative collision avoidance control between each vehicle; a decision module, which is used to establish an AI-based decision model, perform real-time analysis of the fleet status through a machine learning algorithm, and set a reward function for optimizing vehicle control; a lane change control module, which is used to perform multi-vehicle collaborative lane change control in emergency situations; a trajectory tracking module, which is used to adopt a model predictive control strategy to track the trajectory of each vehicle.

[0101] The computer program product of the AI-based multi-vehicle driving control method and device provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the previous method embodiments. The specific implementation can be found in the method embodiments, which will not be repeated here.

[0102] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, it can execute the above-mentioned AI-based multi-vehicle driving control method, thereby improving the safety and efficiency of multi-vehicle driving control.

[0103] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that can be executed by a processor. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0104] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A multi-vehicle driving control method based on AI, characterized in that: include: Establishing a two-way communication topology for the fleet and collecting status information of each vehicle in the fleet; A linear superposition strategy is adopted to coordinate the speed and position of each vehicle in the convoy; Adopt distributed control strategy to carry out cooperative collision avoidance control between each vehicle; Establish an AI-based decision-making model, use machine learning algorithms to analyze fleet status in real time, and set reward functions for optimizing vehicle control; In emergency situations, multiple vehicles can be coordinated to change lanes, so that vehicles can move along the planned trajectory; The model predictive control strategy is used to track the trajectory of each vehicle; The linear superposition strategy is used to coordinate the speed and position of each vehicle in the fleet, including: Establish a nonlinear longitudinal dynamic model of the vehicle to describe the relationship between acceleration and various forces ,in, is the speed of the i-th vehicle at time t, is the input acceleration applied to the vehicle, is the actual driving force acting on the wheel, is the mass of the i-th vehicle, is the headwind drag coefficient, is the rolling friction resistance, is the acceleration due to gravity; Set the control target so that the speed of the main vehicle and the pilot vehicle are consistent ,in, The speed of the pilot car, the main car and the front car maintain a safe distance ,in, is the position of the i-1th vehicle, is the position of the i-th vehicle, is the safe distance between the i-1th vehicle and the i-th vehicle; According to the cooperative adaptive cruise control strategy, the desired input acceleration of the i-th vehicle is controlled ,in, is the actual displacement difference between the vehicle in front and the vehicle in front, is the actual displacement difference between the vehicle and the following vehicle, is the ideal following distance between the i-th vehicle and the preceding vehicle, d is the minimum displacement difference in a stationary state, m is the following coefficient, is the distance tracking coefficient, is the speed tracking coefficient, is the speed tracking coefficient relative to the pilot vehicle; According to the cooperative adaptive cruise control strategy, the distance error between the i-th vehicle and the front vehicle is controlled ,in, The ideal safety distance is set.

2. The AI-based multi-vehicle driving control method according to claim 1, characterized in that: The establishing of a two-way communication topology structure for the fleet and collecting status information of each vehicle in the fleet includes: Take each vehicle in the fleet as a node and construct an information neighborhood set for each node i. , the information neighborhood set Contains information about the leading vehicle, the following vehicle, and the pilot vehicle; The status information of each vehicle is collected in real time through the vehicle-mounted sensors, and the status information is periodically sent to the adjacent preceding and following vehicles in standardized data packets through wireless communication technology; Data fusion technology is used to integrate the status information of each vehicle in the fleet to obtain the global fleet status information.

3. The AI-based multi-vehicle driving control method according to claim 1, characterized in that: The distributed control strategy is used to perform the cooperative collision avoidance control between each vehicle, including: Establish a collision time interval indicator to describe the time required for the main vehicle to collide with the front vehicle at the current speed ,in, is the current position of the preceding vehicle, is the current vehicle position, is the length of the current vehicle, is the speed of the front vehicle, is the current vehicle speed; Establish a front vehicle time distance index to describe the time required for the main vehicle to collide with the front vehicle after the front vehicle suddenly brakes ; According to the collision time interval index and the preceding vehicle time distance index, a collision risk index is obtained. ,in, and is the weight factor; Setting a collision risk threshold and a collision risk upper limit, wherein the collision risk threshold is lower than the collision risk upper limit; When the collision risk indicator When the speed is lower than the collision risk threshold, the low-risk area control strategy is adopted to maintain the current speed and continue driving; When the collision risk indicator When the speed is higher than the collision risk threshold and lower than the collision risk upper limit, the medium risk area control strategy is adopted to adjust the vehicle acceleration, reduce the speed difference, and reduce the collision risk index. The value of When the collision risk indicator When the risk exceeds the upper limit of collision risk, the high-risk area emergency collision avoidance strategy is adopted, and the maximum braking force is applied for braking.

4. The AI-based multi-vehicle driving control method according to claim 3 is characterized in that: The high-risk area emergency collision avoidance strategy includes: When the front vehicle brakes urgently, minimize the relative kinetic energy of the convoy, perform front-end collision avoidance, and calculate the total kinetic energy of the convoy , using the objective function Total kinetic energy of the queue The optimization constraints are , , where j is all other vehicles compared with the i-th vehicle, is the mass of the i-th vehicle, is the speed of the i-th vehicle, is the speed of the jth vehicle, is the acceleration of the i-th vehicle, is the maximum braking deceleration of the i-th vehicle, is the maximum speed of the i-th vehicle; When the rear vehicle approaches sharply, the main vehicle accelerates in coordination to avoid rear-end collision. When the distance between the rear vehicle and the main vehicle is less than the minimum warning distance, the main vehicle brakes at the maximum deceleration. When the leading vehicle is cruising at a constant speed, if a collision is still unavoidable, the main vehicle needs to accelerate in coordination, and the additional acceleration is ,in, is the expected acceleration of the following vehicle, The maximum deceleration allowed for the following vehicle.

5. The AI-based multi-vehicle driving control method according to claim 1, characterized in that: The AI-based decision model is established to analyze the fleet status in real time through a machine learning algorithm, and the reward function for optimizing vehicle control is set, including: Establishing a decision model based on machine learning, wherein the input of the decision model is the current state of the fleet, and the output is the corresponding probability distribution; For the current state of the fleet, a corresponding state space vector and action space vector are designed, wherein the state space vector describes the state of the vehicle and its surrounding environment, and the action space vector describes the action performed by the vehicle; Setting a reward function for evaluating the effect of each of the action space vectors, the reward function comprising a safety reward, a comfort reward, an efficiency reward, a cooperation reward, and an environmental impact reward; For the decision model, the selected action space vector is applied to the vehicle, the vehicle state is updated, and the reward obtained is calculated in real time through a reward function based on the updated state and action of the vehicle at each time step, and the calculated reward is used to update the weight of the decision model.

6. The AI-based multi-vehicle driving control method according to claim 5, characterized in that: The safety bonus includes a safety distance bonus and collision risk penalty ,in, is the distance between vehicle i and vehicle j, is the minimum safe distance, and is the weight factor, is the collision risk factor, is the penalty weight; The comfort bonus includes an acceleration smoothness bonus and braking and acceleration comfort bonus ,in, is the actual acceleration of the vehicle, is the ideal acceleration, is the stability weight, is the maximum allowable braking deceleration; The efficiency bonus includes a speed maintenance bonus and energy efficiency penalty ,in, For the ideal speed, is the efficiency weight, is the energy consumption of vehicle i, is the energy consumption penalty weight; The collaboration reward includes information sharing reward and collaborative behavior rewards ,in, is the information sharing degree between vehicle i and vehicle j, is the information sharing weight, is the weight of collaborative capability; The environmental impact incentives include emission control penalties ,in, is the emission of vehicle i, is the emission penalty weight.

7. The AI-based multi-vehicle driving control method according to claim 1, characterized in that: The multi-vehicle coordinated lane changing control in an emergency situation so that the vehicles travel along the planned trajectory includes: The sensor obtains the current convoy's environmental information and generates a lane change intention if a vehicle fails to reach the expected speed or there is an obstacle ahead. Evaluate the status of vehicles in adjacent lanes and determine the feasibility of lane change; If it is determined that lane changing is feasible, trajectory planning is performed to generate a lane changing trajectory so that the vehicle can travel along the planned trajectory.

8. The AI-based multi-vehicle driving control method according to claim 7, characterized in that: The model predictive control strategy is used to track the trajectory of each vehicle, including: In the fleet, a dynamic model is established to predict the state of each vehicle ； ， in, is the state variable of the i-th vehicle, ( ) is the position of the vehicle, For speed, is the heading angle, T is the control period, is the control input of the i-th vehicle, is the output variable of the i-th vehicle, is the output function of the vehicle; In each control cycle T, the dynamic model is used to track the trajectory of each vehicle, and at each time k, the time step of the specified time range is predicted. The vehicle state trajectory within is ; Construct an objective function for minimizing the error between the vehicle state trajectory and the desired trajectory ,in, is the actual output at time k+i, is the expected output at time k+i, Q is the weighting matrix used to control the output error weight, is the amount of change used to control the input, and R is the weighting matrix used to control the weight of the input change; In each control cycle T, the value of the objective function is minimized to solve the optimization problem. ,in, is the control input vector, is the minimum value limit of the control input, It is the maximum value limit of the control input; Use the first value of the optimized control sequence Update the kinetic model ; Measure the actual status data of the vehicle , calculate the error between the predicted state and the measured state ; Use the predicted state of the next step of the state error to correct it and get the corrected predicted state vector .

9. An AI-based multi-vehicle driving control system, characterized in that: include: A communication module, used to establish a two-way communication topology for the fleet and collect status information of each vehicle in the fleet; A linear control module, used for adopting a linear superposition strategy to coordinate the speed and position of each vehicle in the fleet; A distributed control module is used to adopt a distributed control strategy to perform cooperative collision avoidance control between each vehicle; The decision-making module is used to establish an AI-based decision-making model, conduct real-time analysis of the fleet status through machine learning algorithms, and set reward functions for optimizing vehicle control; Lane-changing control module, used for multi-vehicle coordinated lane-changing control in emergency situations; The trajectory tracking module is used to track the trajectory of each vehicle using a model predictive control strategy; The operations performed by the linear control module include: Establish a nonlinear longitudinal dynamic model of the vehicle to describe the relationship between acceleration and various forces ,in, is the speed of the i-th vehicle at time t, is the input acceleration applied to the vehicle, is the actual driving force acting on the wheel, is the mass of the i-th vehicle, is the headwind drag coefficient, is the rolling friction resistance, is the acceleration due to gravity; Set the control target so that the speed of the main vehicle and the pilot vehicle are consistent ,in, The speed of the pilot car, the main car and the front car maintain a safe distance ,in, is the position of the i-1th vehicle, is the position of the i-th vehicle, is the safe distance between the i-1th vehicle and the i-th vehicle; According to the cooperative adaptive cruise control strategy, the desired input acceleration of the i-th vehicle is controlled ,in, is the actual displacement difference between the vehicle in front and the vehicle in front, is the actual displacement difference between the vehicle and the following vehicle, is the ideal following distance between the i-th vehicle and the preceding vehicle, d is the minimum displacement difference in a stationary state, m is the following coefficient, is the distance tracking coefficient, is the speed tracking coefficient, is the speed tracking coefficient relative to the pilot vehicle; According to the cooperative adaptive cruise control strategy, the distance error between the i-th vehicle and the front vehicle is controlled ,in, The ideal safety distance is set.

Images