Intelligent vehicle lane changing decision-making method based on master-slave game in mixed driving scene

Abstract

The invention provides an intelligent vehicle lane change decision-making method based on a master-slave game in a mixed driving scene, which comprehensively considers the human driver style and the dynamic trajectory influence of surrounding vehicles in various typical lane change scenes, and establishes a more comprehensive multi-vehicle game framework based on the master-slave game. Comprehensive game cost and revenue functions are respectively established for intelligent vehicles and human-driven vehicles on adjacent lanes, and an optimal intelligent vehicle acceleration sequence and lane changing time are iteratively calculated through a double-layer optimization model structure. The accurate planning of the lane changing track of the intelligent vehicle can be realized on the premise of considering multiple aspects such as safety, comfort and traffic efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent driving and traffic control technology, and specifically relates to an intelligent vehicle lane-changing decision method based on master-slave game theory to achieve higher safety, comfort and driving efficiency. Background Art

[0002] Autonomous lane changing is a key function of intelligent self-driving vehicles. The rationality of its lane-changing decisions determines the safety of connected vehicles and also has a certain impact on accident rates and traffic congestion. Because the lane-changing scenarios vehicles face are extremely complex and the types and driving behavior characteristics of other interacting vehicles vary, how to coordinate all parties to achieve safe and efficient autonomous lane changes is currently a challenge in this field. Existing lane-changing decision-making methods mainly include rule-based, learning-based, and game-theory-based approaches. However, these existing technologies still have many shortcomings that need to be overcome. For example, rule-based methods are highly limited when applied in complex and non-standard driving environments; learning-based methods rely on large amounts of driving data and computational overhead; and game-theory-based methods lack comprehensive consideration and modeling of the behavior of pedestrians and vehicles other than the vehicle itself, resulting in unsatisfactory decision-making results. Therefore, how to overcome the shortcomings of existing technologies and realize lane-changing decision-making methods that combine safety, comfort, and traffic efficiency is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0003] In view of this, and to address the technical problems existing in this field, the present invention provides a lane-changing decision method for an intelligent vehicle based on a master-slave game in a mixed driving scenario, which specifically includes the following steps:

[0004] Step 1: Define various typical lane-changing scenarios and build a master-slave game framework based on the interactive relationship between the intelligent vehicle as the leader and the human-driven vehicle in the target lane as the follower.

[0005] Step 2: Build an acceleration model for human-driven vehicles during lane changes, and establish a speed decision metric for human-driven vehicles. Based on this, consider driving safety, space, comfort, and traffic efficiency to establish corresponding reward functions. These reward functions are then weighted and integrated to create a comprehensive game reward function for human-driven vehicles corresponding to different driving styles.

[0006] Step 3: Based on the spatial position relationship and relative speed between the intelligent vehicle and surrounding vehicles, cost functions corresponding to driving safety, comfort, and traffic efficiency are established respectively. The cost functions are then weighted and integrated to establish a comprehensive game cost function for intelligent vehicle lane change.

[0007] Step 4: Consider the static game between different intelligent vehicles and introduce the trajectory prediction of surrounding vehicles into the calculation of the comprehensive game benefits of human-driven vehicles and the comprehensive game costs of intelligent vehicles' lane changes, thereby establishing a multi-vehicle game framework for mixed-vehicle scenarios. This multi-vehicle game framework has a two-layer optimization model structure, in which the upper-layer model comprehensively considers the safety, comfort, and traffic efficiency of intelligent vehicles as optimization objectives, and obtains the optimal lane-changing strategy by sequentially solving the acceleration sequence of intelligent vehicles and the comprehensive game costs of lane changes. The lower-layer model receives the optimal lane-changing strategy provided by the upper-layer model, sequentially solves the comprehensive game benefits of human-driven vehicles and their optimal target speed, and feeds the optimal target speed and vehicle status back to the upper-layer model.

[0008] Step 5: Using the status information of the intelligent vehicle and its surrounding human-driven vehicles and other intelligent vehicles, combined with the multi-vehicle game framework established in step 4, iterative calculation is performed to update the output of the intelligent vehicle's lane change time.

[0009] Furthermore, in step 2, for a human-driven vehicle, the following acceleration model is specifically adopted, in which the acceleration decreases as the speed increases:

[0010] a x =β(v target -v init )e -βt

[0011] Among them, a x is the longitudinal acceleration, v target is the target speed, v init is the velocity at the initial moment, t is the time, and β is the slope of the acceleration curve;

[0012] After integrating the acceleration model, we can obtain the velocity and displacement expressions of the human-driven vehicle:

[0013] v=v target -(v target -v init )e -βt

[0014] x=x init +v target t-(v target -v init )(1-e -βt ) / β

[0015] Among them, v is the current vehicle speed, x is the current longitudinal position, and x init is the longitudinal position at the initial moment;

[0016] On this basis, the speed decision quantity of human driving vehicle is defined as S HV ={v target}.

[0017] Furthermore, the following safety benefit function is established for human-driven vehicles considering safety:

[0018]

[0019] Where τ represents the reaction time, tlc represents the time required for lane change, and are the longitudinal speeds of the human-driven vehicle HV and the intelligent vehicle EV at the time tlc when the lane change ends, Th(tlc) represents the headway at the time when the lane change ends, TTC(tlc) represents the time to collision at the time when the lane change ends, and a * Indicates the acceleration during braking;

[0020] Considering the driving space, the following position advantage benefit function is established

[0021]

[0022] Where Δx represents the longitudinal relative position between the human-driven vehicle and the intelligent vehicle at the end of the lane change, and Δx lim is a constant positively correlated with reaction time;

[0023] Considering comfort, the following comfort benefit function is established

[0024]

[0025] Among them, a HV The acceleration selected for the human driver, a max is the maximum permissible acceleration;

[0026] Considering the traffic efficiency, the following traffic efficiency benefit function is established:

[0027]

[0028] Among them, v HV is the speed of human driving vehicle, Δv max is its maximum speed deviation threshold;

[0029] The weighted fusion of each benefit function yields the following comprehensive game benefit function J corresponding to different driving styles of human driving vehicles: HV :

[0030]

[0031] Where, and are the weight factors of each income.

[0032] Furthermore, in step 3, the following safety cost function is established for the intelligent vehicle considering its relationship with the preceding vehicle FV and the human-driven vehicle HV behind in the adjacent lane:

[0033]

[0034] in, and They represent the longitudinal driving safety cost and the lateral driving safety cost respectively;

[0035] The longitudinal driving safety cost is specifically expressed as:

[0036] C safe_log =k v_log λ log (Δv x_log ) 2 +k s_log / [(Δs x_log ) 2 +ε]

[0037]

[0038] The lateral driving safety cost is specifically expressed as:

[0039] C safe_lat =k v_lat λ lat (Δv x_lat ) 2 +k s_lat / [(Δs x_lat ) 2 +ε]

[0040]

[0041] in, and denote the longitudinal speeds of EV and RV respectively, (x RV ,y RV ) and (x EV ,y EV ) are the positions of RV and EV, k v_lat and k s_lat are weighting coefficients, ε is a very small value used to avoid the appearance of zero denominator in the formula, l v To consider the safety factor of the vehicle length;

[0042] Considering comfort, the following comfort cost function is established

[0043]

[0044] Where aEV represents the acceleration of the intelligent vehicle, Δa EV Indicates its increment;

[0045] Considering the traffic efficiency, the following traffic efficiency cost function is established

[0046]

[0047] Among them, v EV is the speed of the smart vehicle, v desire for its expected speed;

[0048] The above cost functions are weighted and integrated to establish the following intelligent vehicle lane-changing comprehensive game cost function:

[0049]

[0050] Where, and are the weight coefficients of each cost respectively.

[0051] Furthermore, after introducing the predicted trajectory of surrounding vehicles in step 4, the comprehensive game cost function of intelligent vehicle lane change is modified into the following form:

[0052]

[0053] Where C leader (t) is the comprehensive game cost of the smart vehicle EV at time t, To predict the comprehensive game cost, To predict the safety cost, To predict the comfort cost, is the predicted traffic efficiency cost, τ is the prediction time interval, k is the number of predicted time frames, is the state information of the smart vehicle EV at time t, is the state information of the human-driven vehicle HV at time t+kτ, where the predicted value is used instead of the current value;

[0054] The comprehensive game profit function of human-driven vehicles is modified into the following form:

[0055]

[0056] Where, J follower (t) is the comprehensive game benefit of human driving vehicles at time t, J follower (state t ,t+kτ) is the predicted comprehensive game benefit, To predict safety benefits, To predict the position advantage benefits, To predict the comfort gain, To predict driving efficiency gains, state t is the state of the human-driven vehicle HV at time t, and ξ is the attenuation factor.

[0057] Furthermore, in step 5, a genetic algorithm is executed in the lower model, using the state of the human driving the vehicle behind the intelligent vehicle as the population, and calculating the individual fitness based on its comprehensive game payoff. The obtained optimal target speed and vehicle state are fed back to the upper model.

[0058] The upper model uses the received optimal target speed and vehicle state of the human-driven vehicle to solve the following optimization objectives:

[0059]

[0060] Where, the subscript L represents the leader, i.e., the intelligent vehicle; F represents the follower, i.e., the human-driven vehicle; v and a represent the corresponding vehicle speed and acceleration, respectively; Q1, Q2, and Q3 are weight matrices. are the distances between the intelligent vehicle and the human-driven vehicle and the surrounding vehicles at time t when the k-step prediction is made, d safe The minimum safety distance should be set so that the safety distance between the master and slave vehicles is greater than the minimum safety distance. The desired speed of the intelligent vehicle should be kept close to the desired speed during the game.

[0061] Get the optimal acceleration control strategy of the intelligent vehicle at time t~t+N-1 {a *,0|t ,…,a *,k|t ,…,a *,N-1|t}; On this basis, the comprehensive game cost of the original lane of the intelligent vehicle is calculated and the comprehensive game cost of lane changing in Represents the comprehensive game value of the lane where the smart vehicle is located, Represents the comprehensive game cost of the intelligent vehicle's adjacent target lanes;

[0062] If the comprehensive cost of the two games is minimized, Determine the lane change time And update the intelligent vehicle strategy set Contains acceleration sequence and lane change flag.

[0063] The master-slave game-based lane-changing decision-making method for intelligent vehicles in mixed-driving scenarios, as provided by the present invention, comprehensively considers the influence of human driver style and the dynamic trajectories of surrounding vehicles in various typical lane-changing scenarios, establishing a comprehensive multi-vehicle game framework based on master-slave game. Comprehensive game cost and reward functions are established for the intelligent vehicle and human-driven vehicles in adjacent lanes. A two-layer optimization model structure is used to iteratively calculate the optimal intelligent vehicle acceleration sequence and lane-changing timing. This method enables precise planning of intelligent vehicle lane-changing trajectories while balancing safety, comfort, and traffic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 Several typical intelligent vehicle lane-changing game scenarios;

[0065] Figure 2 The multi-car game framework structure established in the method provided by the present invention;

[0066] Figure 3 It is a master-slave game decision-making two-layer optimization model structure based on the genetic algorithm of the present invention. DETAILED DESCRIPTION

[0067] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0068] The intelligent vehicle lane-changing decision method based on master-slave game in a mixed driving scenario provided by the present invention specifically includes the following steps:

[0069] Step 1: Define various typical lane-changing scenarios, including but not limited to Figure 1 The three scenarios shown in the figure are: the EV is the intelligent vehicle, the RV is the vehicle behind it, and the FV is the vehicle in front. A master-slave game framework is constructed based on the interactive relationship between the intelligent vehicle as the leader and the human-driven vehicle in the target lane, the RV, as the follower. During the lane change process, the leading vehicle may cut in front of the intelligent vehicle's current lane or remain in the same lane.

[0070] Step 2: Build an acceleration model for human-driven vehicles during lane changes, and establish a speed decision metric for human-driven vehicles. Based on this, consider driving safety, space, comfort, and traffic efficiency to establish corresponding reward functions. These reward functions are then weighted and integrated to create a comprehensive game reward function for human-driven vehicles corresponding to different driving styles.

[0071] Step 3: Based on the spatial position relationship and relative speed between the intelligent vehicle and surrounding vehicles, cost functions corresponding to driving safety, comfort, and traffic efficiency are established respectively. The cost functions are then weighted and integrated to establish a comprehensive game cost function for intelligent vehicle lane change.

[0072] Step 4: Consider the static game between different intelligent vehicles and introduce the trajectory prediction of surrounding vehicles into the calculation of the comprehensive game benefits of human driving vehicles and the comprehensive game cost of intelligent vehicle lane change, thereby establishing Figure 2 The multi-vehicle game framework for a mixed-car scenario is shown in the figure. This multi-vehicle game framework has a two-layer optimization model structure, in which the upper-layer model comprehensively considers the safety, comfort, and traffic efficiency of intelligent vehicles as optimization objectives, and obtains the optimal lane-changing strategy by sequentially solving the acceleration sequence of intelligent vehicles and the comprehensive lane-changing game cost. The lower-layer model receives the optimal lane-changing strategy provided by the upper-layer model, sequentially solves the comprehensive game payoff of human-driven vehicles and their optimal target speed, and feeds the optimal target speed and vehicle status back to the upper-layer model.

[0073] Step 5: Using the status information of the intelligent vehicle and its surrounding human-driven vehicles and other intelligent vehicles, combined with the multi-vehicle game framework established in step 4, iterative calculation is performed to update the output of the intelligent vehicle's lane change time.

[0074] In a preferred embodiment of the present invention, in step 2, the following acceleration model is specifically adopted for a human-driven vehicle, in which the acceleration decreases as the speed increases:

[0075] a x =β(v target -v init )e -βt

[0076] Among them, a x is the longitudinal acceleration, v target is the target speed, v init is the velocity at the initial moment, t is the time, and β is the slope of the acceleration curve;

[0077] After integrating the acceleration model, we can obtain the velocity and displacement expressions of the human-driven vehicle:

[0078] v=v target -(v target -v init )e -βt

[0079] x=x init +v target t-(v target -v init )(1-e -βt ) / β

[0080] Among them, v is the current vehicle speed, x is the current longitudinal position, and x init is the longitudinal position at the initial moment;

[0081] Just know v init 、x init 、v target That is, it can calculate the motion state and behavior of human driving vehicles, and define different v according to different driving styles. target When the autonomous vehicle EV intends to change lanes, the general or conservative driver of the vehicle behind in the target lane expects to slow down or maintain the speed, keep a large distance from the EV, avoid collision and ensure safety; the aggressive driver expects a higher V in the target lane. target To accelerate to prevent EV lane change

[0082] On this basis, the speed decision quantity of human driving vehicle is defined as S HV ={v target}.

[0083] In a preferred embodiment of the present invention, the following safety benefit function is established for human-driven vehicles considering the safety of collision time and headway:

[0084]

[0085] Where τ represents the reaction time, tlc represents the time required for lane change, and are the longitudinal speeds of the human-driven vehicle HV and the intelligent vehicle EV at the time tlc when the lane change ends, Th(tlc) represents the headway at the time when the lane change ends, TTC(tlc) represents the time to collision at the time when the lane change ends, and a * Indicates the acceleration during braking;

[0086] Considering the driving space, the following position advantage benefit function is established

[0087]

[0088] Where Δx represents the longitudinal relative position between the human-driven vehicle and the intelligent vehicle at the end of the lane change, and Δx lim is a constant positively correlated with reaction time;

[0089] Considering comfort, the following comfort benefit function is established

[0090]

[0091] Among them, a HV The acceleration selected for the human driver, amax is the maximum permissible acceleration;

[0092] Considering the traffic efficiency, the following traffic efficiency benefit function is established:

[0093]

[0094] Among them, v HV is the speed of human driving vehicle, Δv max is its maximum speed deviation threshold;

[0095] The weighted fusion of each benefit function yields the following comprehensive game benefit function J corresponding to different driving styles of human driving vehicles: HV :

[0096]

[0097] Where, and are the weight factors of each income.

[0098] In a preferred embodiment of the present invention, in step 3, the following safety cost function is established for the intelligent vehicle, specifically considering its relationship with the preceding vehicle FV and the human-driven vehicle HV behind the adjacent lane:

[0099]

[0100] in, and They represent the longitudinal driving safety cost and the lateral driving safety cost respectively;

[0101] The longitudinal driving safety cost is specifically expressed as:

[0102] C safe_log =k v_log λ log (Δv x_log ) 2 +k s_log / [(Δs x_log ) 2 +ε]

[0103]

[0104] The lateral driving safety cost is specifically expressed as:

[0105] C safe_lat =k v_lat λ lat (Δv x_lat ) 2 +k s_lat / [(Δs x_lat )2 +ε]

[0106]

[0107] in, and denote the longitudinal speeds of EV and RV respectively, (x RV ,y RV ) and (x EV ,y EV ) are the positions of RV and EV, k v_lat and k s_lat are weighting coefficients, ε is a very small value used to avoid the appearance of zero denominator in the formula, l v To consider the safety factor of the vehicle length;

[0108] Considering comfort, the following comfort cost function is established

[0109]

[0110] Where a EV represents the acceleration of the intelligent vehicle, Δa EV Indicates its increment;

[0111] Considering the traffic efficiency, the following traffic efficiency cost function is established

[0112]

[0113] Among them, v EV is the speed of the smart vehicle, v desire for its expected speed;

[0114] The above cost functions are weighted and integrated to establish the following intelligent vehicle lane-changing comprehensive game cost function:

[0115]

[0116] Where, and are the weight coefficients of each cost respectively.

[0117] In a preferred embodiment of the present invention, after the predicted trajectories of surrounding vehicles are introduced in step 4, the comprehensive game cost function of the intelligent vehicle lane change is modified to the following form:

[0118]

[0119] Where C leader (t) is the comprehensive game cost of the smart vehicle EV at time t, To predict the comprehensive game cost, To predict the safety cost, To predict the comfort cost, is the predicted traffic efficiency cost, τ is the prediction time interval, k is the number of predicted time frames, is the state information of the smart vehicle EV at time t, is the state information of the human-driven vehicle HV at time t+kτ, where the predicted value is used instead of the current value;

[0120] The comprehensive game profit function of human-driven vehicles is modified into the following form:

[0121]

[0122] Where, J follower (t) is the comprehensive game benefit of human driving vehicles at time t, J follower (state t ,t+kτ) is the predicted comprehensive game benefit, To predict safety benefits, To predict the position advantage benefits, To predict the comfort gain, To predict driving efficiency gains, state t is the state of the human-driven vehicle HV at time t, and ξ is the attenuation factor.

[0123] In a preferred embodiment of the present invention, in step 5, Figure 3 As shown in the figure, the lower model executes a genetic algorithm, takes the state of human-driven vehicles behind the intelligent vehicle as the population, calculates individual fitness based on its comprehensive game benefits, and feeds the obtained optimal target speed and vehicle state back to the upper model;

[0124] The upper model uses the received optimal target speed and vehicle state of the human-driven vehicle to solve the following optimization objectives:

[0125]

[0126] Where, the subscript L represents the leader, i.e., the intelligent vehicle; F represents the follower, i.e., the human-driven vehicle; v and a represent the corresponding vehicle speed and acceleration, respectively; Q1, Q2, and Q3 are weight matrices. are the distances between the intelligent vehicle and the human-driven vehicle and the surrounding vehicles at time t when the k-step prediction is made, d safe The minimum safety distance should be set so that the safety distance between the master and slave vehicles is greater than the minimum safety distance. is the desired speed of the intelligent vehicle, and the speed should be close to the desired speed during the game process; the optimal acceleration control strategy of the intelligent vehicle at time t~t+N-1 is obtained {a *,0|t ,…,a*,k|t ,…,a *,N-1|t};

[0127] On this basis, the comprehensive game cost of the original lane of the intelligent vehicle is calculated and the comprehensive game cost of lane changing in Represents the comprehensive game value of the lane where the smart vehicle is located, Represents the comprehensive game cost of the intelligent vehicle's adjacent target lanes;

[0128] If the comprehensive cost of the two games is minimized, Determine the lane change time And update the intelligent vehicle strategy set Contains acceleration sequence and lane change flag.

[0129] It should be understood that the size of the serial numbers of the steps in the embodiment of the present invention does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0130] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent vehicle lane-changing decision-making method based on master-slave game in mixed driving scenarios, characterized by: The specific steps include: Step 1: Define various typical lane-changing scenarios and build a master-slave game framework based on the interactive relationship between the intelligent vehicle as the leader and the human-driven vehicle in the target lane as the follower. Step 2: Build an acceleration model for human-driven vehicles during lane changes, and establish a speed decision metric for human-driven vehicles. Based on this, consider driving safety, space, comfort, and traffic efficiency to establish corresponding reward functions. These reward functions are then weighted and integrated to create a comprehensive game reward function for human-driven vehicles corresponding to different driving styles. Step 3: Based on the spatial position relationship and relative speed between the intelligent vehicle and surrounding vehicles, cost functions corresponding to driving safety, comfort, and traffic efficiency are established respectively. The cost functions are then weighted and integrated to establish a comprehensive game cost function for intelligent vehicle lane change. Step 4: Consider the static game between different intelligent vehicles and incorporate the trajectory prediction of surrounding vehicles into the calculation of the comprehensive game benefits of human-driven vehicles and the comprehensive game costs of lane changes for intelligent vehicles. This establishes a multi-vehicle game framework for mixed-vehicle scenarios. This multi-vehicle game framework has a two-layer optimization model structure. The upper layer comprehensively considers the safety, comfort, and traffic efficiency of intelligent vehicles as optimization objectives. The optimal lane-changing strategy is obtained by sequentially solving the acceleration sequence of intelligent vehicles and the comprehensive game costs of lane changes. The lower model receives the optimal lane-changing strategy from the upper model, solves the comprehensive game payoff of the human-driven vehicle and its optimal target speed, and feeds the optimal target speed and vehicle status back to the upper model. Step 5: Using the status information of the intelligent vehicle and its surrounding human-driven vehicles and other intelligent vehicles, combined with the multi-vehicle game framework established in step 4, iterative calculation is performed to update the output of the intelligent vehicle's lane change time.

2. The method according to claim 1, wherein: In step 2, the following acceleration model is used for the human-driven vehicle, where the acceleration decreases as the speed increases: a x =β(v target -v init )e -βt Among them, a x is the longitudinal acceleration, v target is the target speed, v init is the velocity at the initial moment, t is the time, and β is the slope of the acceleration curve; After integrating the acceleration model, we can obtain the velocity and displacement expressions of the human-driven vehicle: v=v target -(v target -v init )e -βt x=x init +v target t-(v target -v init )(1-e -βt ) / β Among them, v is the current vehicle speed, x is the current longitudinal position, and x init is the longitudinal position at the initial moment; On this basis, the speed decision quantity of human driving vehicle is defined as S HV ={v target }.

3. The method according to claim 2, wherein: Considering the safety of human-driven vehicles, the following safety benefit function is established Where τ represents the reaction time, tlc represents the time required for lane change, and are the longitudinal speeds of the human-driven vehicle HV and the intelligent vehicle EV at the time tlc when the lane change ends, Th(tlc) represents the headway at the time when the lane change ends, TTC(tlc) represents the time to collision at the time when the lane change ends, and a * Indicates the acceleration during braking; Considering the driving space, the following position advantage benefit function is established Where Δx represents the longitudinal relative position between the human-driven vehicle and the intelligent vehicle at the end of the lane change, and Δx lim is a constant positively correlated with reaction time; Considering comfort, the following comfort benefit function is established Among them, a HV The acceleration selected for the human driver, a max is the maximum permissible acceleration; Considering the traffic efficiency, the following traffic efficiency benefit function is established: Among them, v HV is the speed of human driving vehicle, Δv max is its maximum speed deviation threshold; The weighted fusion of each benefit function yields the following comprehensive game benefit function J corresponding to different driving styles of human driving vehicles: HV : Where, and are the weight factors of each income.

4. The method according to claim 3, wherein: In step 3, the following safety cost function is established for the intelligent vehicle considering its relationship with the preceding vehicle FV and the human-driven vehicle HV behind the adjacent lane: in, and They represent the longitudinal driving safety cost and the lateral driving safety cost respectively, and σ is the lane change sign; The longitudinal driving safety cost is specifically expressed as: C safe_log =k v_log l log (Δv x_log ) 2 +k s_log / [(Δs x_log ) 2 +e] The lateral driving safety cost is specifically expressed as: in, and denote the longitudinal speeds of EV and RV respectively, (x RV ,y RV ) and (x EV ,y EV ) are the positions of RV and EV, k v_lat and k s_lat are weighting coefficients, ε is a very small value used to avoid the appearance of zero denominator in the formula, l v To consider the safety factor of the vehicle length; Considering comfort, the following comfort cost function is established Where a EV represents the acceleration of the intelligent vehicle, Δa EV Indicates its increment; Considering the traffic efficiency, the following traffic efficiency cost function is established Among them, v EV is the speed of the smart vehicle, v desire for its expected speed; The above cost functions are weighted and integrated to establish the following intelligent vehicle lane-changing comprehensive game cost function: Where, and are the weight coefficients of each cost respectively.

5. The method according to claim 4, wherein: After introducing the predicted trajectories of surrounding vehicles in step 4, the comprehensive game cost function of intelligent vehicle lane change is modified to the following form: Where C leader (t) is the comprehensive game cost of the smart vehicle EV at time t, To predict the comprehensive game cost, To predict the safety cost, To predict the comfort cost, is the predicted traffic efficiency cost, τ is the prediction time interval, k is the number of predicted time frames, is the state information of the smart vehicle EV at time t, is the state information of the human-driven vehicle HV at time t+kτ, where the predicted value is used instead of the current value; The comprehensive game profit function of human-driven vehicles is modified into the following form: Where, J follower (t) is the comprehensive game benefit of human driving vehicles at time t, J follower (state t ,t+kτ) is the predicted comprehensive game benefit, To predict safety benefits, To predict the position advantage benefits, To predict the comfort gain, To predict driving efficiency gains, state t is the state of the human-driven vehicle HV at time t, and ξ is the attenuation factor.

6. The method according to claim 5, wherein: In step 5, a genetic algorithm is executed in the lower-level model, using the state of the human driving the vehicle behind the intelligent vehicle as the population. The individual fitness is calculated based on their comprehensive game benefits, and the obtained optimal target speed and vehicle state are fed back to the upper-level model. The upper model uses the received optimal target speed and vehicle state of the human-driven vehicle to solve the following optimization objectives: Where, the subscript L represents the leader, i.e., the intelligent vehicle; F represents the follower, i.e., the human-driven vehicle; v and a represent the corresponding vehicle speed and acceleration, respectively; Q1, Q2, and Q3 are weight matrices. are the distances between the intelligent vehicle and the human-driven vehicle and the surrounding vehicles at time t when the k-step prediction is made, d safe The minimum safety distance should be set so that the safety distance between the master and slave vehicles is greater than the minimum safety distance. The desired speed of the intelligent vehicle should be kept close to the desired speed during the game. Get the optimal acceleration control strategy of the intelligent vehicle at time t~t+N-1 {a *,0|t ,…,a *,k|t ,…,a *,N-1|t }; On this basis, the comprehensive game cost of the original lane of the intelligent vehicle is calculated and the comprehensive game cost of lane changing in Represents the comprehensive game value of the lane where the smart vehicle is located, Represents the comprehensive game cost of the intelligent vehicle's adjacent target lanes; If the comprehensive cost of the two games is minimized, Determine the lane change time And update the intelligent vehicle strategy set Contains acceleration sequence and lane change flag.

Images