Evolutionary game decision-making method for lane changing of autonomous vehicle

Abstract

The invention belongs to the technical field of lane changing decision making, and particularly relates to an evolutionary game decision making method for lane changing of an automatic driving vehicle, which comprises the following steps: in the process of changing the automatic driving vehicle from a current lane to a target lane, pre-judging a target vehicle of the automatic driving vehicle, collecting motion information of the automatic driving lane changing vehicle and the target vehicle in real time, and determining a game scene of lane changing of the automatic driving lane changing vehicle; defining an automatic driving lane changing vehicle and a target vehicle as decision subjects, and determining a strategy set of the automatic driving lane changing vehicle and the target vehicle and a strategy combination of game of the two; constructing an evolutionary game model of interactive decision of the automatic driving lane changing vehicle and the target vehicle; establishing a replication dynamic equation of the decision subject; and according to an optimization result obtained by the evolutionary game model, making a lane changing decision and executing the lane changing decision. According to the invention, the automatic driving vehicle can adjust the lane changing strategy in real time according to the traffic environment, so that the performance of the man-machine co-driving system is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of lane changing decision making, and in particular relates to an evolutionary game decision making method for lane changing of an autonomous driving vehicle. Background Art

[0002] With the development of science and technology, the development and popularization of autonomous vehicles are expected to reduce traffic accidents, improve vehicle traffic efficiency, ease traffic congestion, and reduce traffic emissions. The Society of Automatic Machine Engineers divides autonomous driving technology into six levels: manual driving, assisted driving, partial autonomous driving, conditional autonomous driving, highly autonomous driving, and fully autonomous driving. At present, the stage of autonomous driving technology is: partial autonomous driving and conditional autonomous driving are mature, and highly autonomous driving has begun testing. Although autonomous driving technology has made many breakthroughs, due to the fact that the technology is not yet fully mature and the laws and regulations are not perfect, before fully autonomous driving is achieved, there must be both manually driven vehicles and autonomous driving vehicles of different levels in the road traffic system.

[0003] Autonomous driving technology mainly consists of three parts: information perception, decision planning, and control execution. Decision planning is the key to ensure that autonomous driving vehicles can accurately and smoothly complete various driving tasks. A very important decision for autonomous driving vehicles is lane changing. Existing studies have proposed a variety of autonomous driving lane changing decision methods based on logical rules, machine learning, and game theory, but they all have shortcomings. The lane changing decision method based on logical rules relies on predefined rules with logical relationships, lacks sufficient flexibility, and is difficult to adapt to complex traffic environments. Machine learning can learn driving behavior from driving behavior data, but it relies on a large amount of high-quality training data and has limited generalization capabilities for unknown situations.

[0004] Classical games in game theory can consider interactive decisions between vehicles, but due to one-time or limited rounds of games, the decision-making process lacks real-time and adaptability. There have been studies on lane-changing decisions for autonomous vehicles based on evolutionary games. When describing vehicle utility, most of them consider three factors: safety, efficiency, and comfort, but ignore two factors: driver fatigue and energy saving of autonomous vehicles, resulting in unreasonable evolutionary game results. Summary of the invention

[0005] In view of the above deficiencies in the prior art, the purpose of the present invention is to provide an evolutionary game decision-making method for lane changing of an autonomous driving vehicle, which integrates multiple benefits of safety, efficiency, comfort, fatigue and energy saving, and enables the autonomous driving vehicle to adjust its lane changing strategy in real time according to the traffic environment, so as to improve the performance of the human-machine co-driving system.

[0006] To achieve the above objectives, the present invention provides an evolutionary game decision-making method for lane changing of an autonomous driving vehicle, comprising the following steps: S1. When the autonomous driving lane-changing vehicle changes lanes from the current lane to the target lane, predict the target vehicle of the autonomous driving lane-changing vehicle, collect the motion information of the autonomous driving lane-changing vehicle and the target vehicle in real time through sensors, and determine the game scenario of the autonomous driving lane-changing vehicle changing lanes; S2. Define the autonomous lane-changing vehicle and the target vehicle as decision-making entities, determine the strategy set of the autonomous lane-changing vehicle and the target vehicle, and the strategy combination of the game between the autonomous lane-changing vehicle and the target vehicle; S3. Construct an evolutionary game model for interactive decision-making between the autonomous lane-changing vehicle and the target vehicle, including the utility function and utility matrix of the autonomous lane-changing vehicle and the target vehicle; S4. Establish a replica dynamic equation of the decision-making subject, simulate the strategy selection process of the decision-making subject, and adjust its strategy selection according to the utility of each strategy; S5. Make and execute lane-changing decisions based on the optimization results obtained from the evolutionary game model.

[0007] As a preferred solution of the present invention, in S1, the motion information includes position, speed, and acceleration.

[0008] As a preferred solution of the present invention, in S1, the game scenario is specifically: On a one-way two-lane road or above, if the utility obtained by the vehicle in the current lane is lower than the utility obtained in the target lane, the vehicle will make a lane-changing decision analysis based on the operating status of itself and surrounding vehicles to decide whether to change lanes to the target lane. At this time, the vehicle in the current lane is the lane-changing vehicle, and the rear vehicle in the target lane is the target vehicle. If the behavior decisions of the lane-changing vehicle and the target vehicle can be accurately predicted, no game occurs between the lane-changing vehicle and the target vehicle; specifically, if the lane-changing vehicle has started to change lanes and occupies the target lane, the target vehicle can only give way, and the lane-changing vehicle continues to complete the lane change. At this time, no game occurs between the lane-changing vehicle and the target vehicle; if the target vehicle has not reached the conflict zone, the lane-changing vehicle performs the lane-changing operation, and the target vehicle does not need to brake or slow down and drives normally. At this time, no game occurs between the lane-changing vehicle and the target vehicle; except for these two situations, a game occurs between the lane-changing vehicle and the target vehicle; Considering that the lane-changing vehicle is an autonomous driving vehicle, and the target vehicle is a manually driven vehicle or an autonomous driving vehicle, the game scenarios of autonomous driving vehicle lane changing are divided into scenario 1 and scenario 2. Scenario 1: the game between the autonomous driving lane-changing vehicle and the manually driven target vehicle, and scenario 2: the game between the autonomous driving lane-changing vehicle and the autonomous driving target vehicle.

[0009] As a preferred solution of the present invention, in S2, the strategy set of the automatic driving lane changing vehicle is {A i|i=1,2}={change lane, do not change lane}, the target vehicle’s strategy set is {T j |j=1,2}={give way, don’t give way}; the strategy combination of the game between the autonomous driving lane-changing vehicle and the target vehicle is {(A i ,T j )|i,j∈{1,2}}={(change lanes, give way),(change lanes, do not give way),(do not change lanes, give way),(do not change lanes, do not give way)}.

[0010] As a preferred embodiment of the present invention, in S3, for , the utility function of the decision-making subject includes: The safety utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: (1); (2); In the formula, , are the safety utilities of the automated lane-changing vehicle and the target vehicle, respectively; d is the distance between the automated lane-changing vehicle and the target vehicle; , are the speeds of the autonomous driving lane-changing vehicle and the target vehicle at time t, respectively; , Represent the logical operators "and" and "or" respectively; The efficiency utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: (3); (4); In the formula, , are the efficiency utilities of the autonomous lane-changing vehicle and the target vehicle, respectively; The desired speeds of the lane-changing vehicle and the target vehicle for automated driving; , are the accelerations of the autonomous driving lane-changing vehicle and the target vehicle at time t, respectively; The comfort utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: (5); (6); In the formula, , are the comfort utilities of the autonomous lane-changing vehicle and the target vehicle, respectively; , are the accelerations of the automated lane-changing vehicle and the target vehicle at time t+4f, respectively, and f is the duration of each frame; , are the maximum acceleration and maximum deceleration of the autonomous lane-changing vehicle and the target vehicle, respectively; The fatigue utility function of the manually driven target vehicle is: (7); In the formula, Fatigue effect of manually driven target vehicle; , are the accelerations of the manually driven target vehicle at time t and time t+4f, respectively; The energy-saving utility functions of the autonomous driving lane-changing vehicle and the autonomous driving target vehicle are: (8); (9); In the formula, , They are the energy-saving utility of the autonomous lane-changing vehicle and the autonomous target vehicle, respectively; and are the frontal area and mass of the autonomous lane-changing vehicle, respectively; and are the frontal area and mass of the autonomous driving target vehicle respectively; is the quality factor; is the road slope; is the air density; is the wind speed; g is the acceleration due to gravity; , , They are rolling resistance coefficient, wind resistance coefficient and internal friction resistance coefficient; Each utility function is normalized, and its normalized value is expressed as: (10); In the formula, is the kth utility value obtained by decision subject N using G when the strategy combination is {i, j}; N={ALV, TV}, ALV and TV are the autonomous lane-changing vehicle and the target vehicle respectively; G={S, E, C, F, ES}, S, E, C, F and ES represent safety utility, efficiency utility, comfort utility, fatigue utility and energy-saving utility respectively; k={1, 2, 3, …, K}, K is the sample size of the utility value obtained by the decision subject; make , , Represent the total utility of the autonomous driving lane-changing vehicle, the total utility of the manually driven target vehicle, and the total utility of the autonomous driving target vehicle, respectively. Then: (11); (12); (13).

[0011] As a preferred solution of the present invention, in S3, the utility matrix of the decision-making subject specifically includes: The utility matrix of the autonomous lane-changing vehicle and the manually driven target vehicle. When the autonomous lane-changing vehicle chooses to change lanes and the manually driven target vehicle chooses to give way, the utility is ; When the autonomous lane-changing vehicle chooses to change lanes and the manually driven target vehicle chooses not to give way, the utility is ; When the autonomous lane-changing vehicle chooses not to change lanes and the manually driven target vehicle chooses to give way, the utility is ; When the autonomous lane-changing vehicle chooses not to change lanes and the manually driven target vehicle chooses not to give way, the utility is ; The utility matrix of the autonomous lane-changing vehicle and the autonomous target vehicle. When the autonomous lane-changing vehicle chooses to change lanes and the autonomous target vehicle chooses to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses to change lanes and the autonomous driving target vehicle chooses not to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses not to change lanes and the autonomous driving target vehicle chooses to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses not to change lanes and the autonomous driving target vehicle chooses not to give way, the utility is ; The proportion of the lane-changing strategy and the non-lane-changing strategy selected by the autonomous driving lane-changing vehicle are recorded as p and 1-p respectively, and the proportion of the target vehicle selecting the yielding strategy and the non-yielding strategy are recorded as q and 1-q respectively.

[0012] As a preferred solution of the present invention, in S4, the replication dynamic equation of the decision-making subject includes: Expected utility of lane-changing strategy and non-lane-changing strategy for autonomous driving lane-changing vehicles , They are: (14); (15); Average expected utility of autonomous lane-changing vehicles choosing lane-changing strategies and non-lane-changing strategies for: (16); Replicated dynamic equations for lane-changing strategy selection by autonomous driving vehicle for: (17); Expected utility of manually driven target vehicles choosing the give way strategy and the no-give way strategy , They are: (18); (19); Average expected utility of manually driven target vehicles choosing the give way strategy and the no-give way strategy for: (20); Replicated dynamic equations for the yield strategy of manually driven target vehicles It is expressed as: (twenty one); Expected utility of autonomous driving target vehicle choosing yielding strategy and not yielding strategy , They are: (twenty two); (twenty three); Average expected utility of the autonomous driving target vehicle choosing the give way strategy and the no-give way strategy for: (twenty four); Replicated dynamic equations for autonomous driving target vehicle's choice of yield strategy for: (25).

[0013] As a preferred solution of the present invention, for the replication dynamic equation of the decision-making subject, let When choosing a yielding strategy or a non-yielding strategy for the target vehicle, the utility of the lane changing strategy and the non-lane changing strategy for the autonomous driving lane changing vehicle is different, that is, in scenarios 1 and 2, , ; make When choosing a lane-changing strategy or a non-lane-changing strategy for an autonomous vehicle, the utility of the target vehicle choosing a yielding strategy or a non-yielding strategy is different, that is, in scenario 1, , In scene 2, , ; Then equation (17) is simplified to equation (26), and equations (21) and (25) are simplified to equation (27): (26); (27); In the formula, represent or .

[0014] As a preferred solution of the present invention, in S5, the solution of the replicated dynamic equation is the optimization result of the evolutionary game model, and a specific lane-changing decision is made based on the result, including the equilibrium point and the stability of the equilibrium point, wherein: For the equilibrium point, the differential equations about p and q composed of equations (26) and (27) are solved to obtain equation (28), which gives the five equilibrium points of the game between the autonomous driving lane-changing vehicle and the target vehicle, namely (0,0), (0,1), (1,0), (1,1) and : (28); The Jacobian matrix J in the game is: (29); Regarding the stability of the equilibrium point, whether the equilibrium point is a stable point, a saddle point, a central point or an unstable point depends on the determinant and the sign of the trace of the Jacobian matrix; the judgment criterion of the equilibrium point type is obtained according to the Friedman method. When the determinant of the Jacobian matrix is ​​greater than 0 and the trace is less than 0, the equilibrium point is a stable point, and the strategy selected by the decision-making subject is an evolutionary stable strategy.

[0015] As a preferred embodiment of the present invention, the Jacobian matrix determinant and the sign of the trace of the evolutionary game of the lane-changing vehicle of the automatic driving lane-changing vehicle depend on , , , The positive and negative Based on the utility function and utility matrix, in scenario 1, The eight situations are as follows: Scenario 1: , , and , the stable point is (1,1); Scenario 2: , , and , the stable points are (0,0) and (1,1); Scenario 3: , , and , the stable point is (0,1); Scenario 4: , , and , the stable point is (0,0); Scenario 5: , , and , no stable point; Scenario 6: , , and , the stable point is (0,0); Scenario 7: , , and , the stable point is (0,1); Scenario 8: , , and , the stable point is (0,0); In scenario 2, and It is always established and meets the conditions 1 to 4; Assuming the initial values ​​of p and q to be 0~1 respectively, the game evolution results are obtained by numerical simulation; according to the stability analysis of the game equilibrium point, the evolutionary stable strategies of the game system include the following three combinations: (lane change, give way), (no lane change, no give way) and (no lane change, give way); Since the strategy combination (lane change, no giving way) will bring safety risks to both decision-makers, the corresponding equilibrium point cannot be a stable point; The autonomous lane-changing vehicle makes behavioral decisions based on the strategy corresponding to the evolutionary game results: first, identify the evolutionarily stable strategy; then, determine whether the currently selected strategy is an evolutionarily stable strategy based on the current operating status. If the currently selected strategy is an evolutionarily stable strategy, then maintain the currently selected strategy; if the currently selected strategy is not an evolutionarily stable strategy, then select the evolutionarily stable strategy; if there are multiple evolutionarily stable strategies, then select the strategy with the largest total utility.

[0016] The algorithm involved in the present invention can be executed by an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the above algorithm calculation is implemented by executing the software by the processor.

[0017] The beneficial effects of the present invention are: The present invention integrates multiple utilities such as safety, efficiency, comfort, fatigue and energy saving to construct an evolutionary game model for interactive decision-making between autonomous driving lane-changing vehicles and target vehicles, uses replicated dynamic equations to simulate the strategy selection process of the decision-making subject, and solves the equilibrium point of the evolutionary game model. The stability of the equilibrium point is determined according to the Friedman method, and then the evolutionary stable strategy is determined. According to changes in the traffic environment, it helps autonomous driving vehicles adjust their lane-changing strategies in real time, improves the safety and intelligence of lane-changing decisions of autonomous driving lane-changing vehicles through strategy optimization, and better adapts to complex traffic environments.

[0018] The utility function of the lane-changing game of the autonomous driving vehicle in the present invention not only considers safety, efficiency and comfort factors, but also fatigue and energy-saving factors. By considering fatigue utility, driving errors caused by driver fatigue are reduced, and the lane-changing strategy of the autonomous driving vehicle is optimized. By considering energy-saving utility, autonomous driving vehicles are encouraged to choose low-energy consumption and low-emission strategies, promoting environmental protection and sustainable development. This multi-utility fusion provides a systematic lane-changing decision framework for autonomous driving vehicles, which can achieve a win-win situation in economic and environmental benefits, and endow the intelligent transportation system with the dual concepts of intelligence and sustainability. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 is a schematic diagram of a non-game scenario between an autonomous driving lane-changing vehicle and a target vehicle in an embodiment of the present invention, Figure 2 (a) is a schematic diagram of an autonomous driving lane-changing vehicle occupying a target lane; Figure 2 (b) is a schematic diagram showing that the target vehicle has not reached the conflict zone; Figure 3 is a schematic diagram of a game scenario of an automatic driving lane-changing vehicle in an embodiment of the present invention; Figure 3 (a) is a schematic diagram of scenario 1; Figure 3 (b) is a schematic diagram of scenario 2; Figure 4 is a schematic diagram of a shooting scene during the verification process of the present invention; Figure 5 is a schematic diagram of determining whether a target vehicle has reached a conflict zone during the verification process of the present invention; Figure 6 It is an evolutionary path diagram of the game between the autonomous driving lane-changing vehicle and the target vehicle during the verification process of the present invention; Figure 6 (a) in the figure is the evolution path diagram with the stable point (0,0); Figure 6 (b) in the figure is the evolution path diagram with the stable point (0,1); Figure 6 (c) in the figure is the evolution path diagram with the stable point (1,1); Figure 6(d) in the figure is the evolution path diagram with stable points (0,0) and (1,1); Figure 7 is an evolutionary path diagram of lane changing of an autonomous driving vehicle during verification of the present invention; Figure 7 (a) in the figure is the evolution path diagram of scenario 1; Figure 7 (b) in the figure is the evolution path diagram of scenario 2. DETAILED DESCRIPTION

[0020] The embodiments of the present invention are further described below in conjunction with the accompanying drawings: like Figure 1 As shown, an evolutionary game decision-making method for lane changing of an autonomous driving vehicle includes the following steps: S1. When the autonomous driving lane-changing vehicle changes lanes from the current lane to the target lane, predict the target vehicle of the autonomous driving lane-changing vehicle, collect the motion information of the autonomous driving lane-changing vehicle and the target vehicle in real time through sensors, and determine the game scenario of the autonomous driving lane-changing vehicle changing lanes; S2. Define the autonomous lane-changing vehicle and the target vehicle as decision-making entities, determine the strategy set of the autonomous lane-changing vehicle and the target vehicle, and the strategy combination of the game between the autonomous lane-changing vehicle and the target vehicle; S3. Construct an evolutionary game model for interactive decision-making between the autonomous lane-changing vehicle and the target vehicle, including the utility function and utility matrix of the autonomous lane-changing vehicle and the target vehicle; S4. Establish a replica dynamic equation of the decision-making subject, simulate the strategy selection process of the decision-making subject, and adjust its strategy selection according to the utility of each strategy; S5. Make and execute lane-changing decisions based on the optimization results obtained from the evolutionary game model.

[0021] Based on the above method, the lane change goal can be effectively achieved under the premise of safety. Motion information includes position, speed, and acceleration.

[0022] In S1, the specific game scenario is: On a one-way two-lane road or above, if the utility obtained by the vehicle in the current lane is lower than the utility obtained in the target lane, the vehicle will make a lane-changing decision analysis based on the operating status of itself and surrounding vehicles to decide whether to change lanes to the target lane. At this time, the vehicle in the current lane is the lane-changing vehicle, and the rear vehicle in the target lane is the target vehicle. If the behavior decisions of the lane-changing vehicle and the target vehicle can be accurately predicted, no game occurs between the lane-changing vehicle and the target vehicle. Specifically, if the lane-changing vehicle has started to change lanes and occupies the target lane, the target vehicle can only give way while the lane-changing vehicle continues to complete the lane change. At this time, no game occurs between the lane-changing vehicle and the target vehicle. If the target vehicle has not reached the conflict zone (the target vehicle is quite far away from the lane-changing vehicle), the lane-changing vehicle performs the lane-changing operation, and the target vehicle does not need to brake or decelerate and drives normally. At this time, no game occurs between the lane-changing vehicle and the target vehicle. Except for these two situations, a game occurs between the lane-changing vehicle and the target vehicle.

[0023] Figure 2 It demonstrates a scenario where the autonomous lane-changing vehicle and the target vehicle do not compete with each other. Figure 2 In (a), the autonomous lane-changing vehicle (ALV) has started to change lanes and occupies the target lane (TL). The target vehicle (TV) can only give way while the autonomous lane-changing vehicle continues to complete the lane change. At this time, there is no game between the autonomous lane-changing vehicle and the target vehicle. Figure 2 In (b), the target vehicle has not reached the conflict zone, the autonomous lane-changing vehicle performs the lane-changing operation, and the target vehicle does not need to brake or slow down and drives normally. At this time, there is no game between the autonomous lane-changing vehicle and the target vehicle. Except for these two situations, there is a game between the autonomous lane-changing vehicle and the target vehicle.

[0024] Considering that the lane-changing vehicle is an autonomous vehicle, and the target vehicle is a manually driven vehicle or an autonomous vehicle, the game scenario of the lane-changing of the autonomous vehicle is divided into scenario 1 and scenario 2. Scenario 1: Figure 3 As shown in (a) of the figure, the autonomous lane-changing vehicle competes with the manually driven target vehicle (HTV). Scenario 2: Figure 3 As shown in (b) in the figure, the autonomous lane-changing vehicle competes with the autonomous target vehicle (ATV).

[0025] In S2, the strategy set of the autonomous driving lane-changing vehicle is {A i |i=1,2}={change lane, do not change lane}, the target vehicle’s strategy set is {T j |j=1,2}={give way, don’t give way}; the strategy combination of the game between the autonomous driving lane-changing vehicle and the target vehicle is {(A i ,T j )|i,j∈{1,2}}={(change lanes, give way),(change lanes, do not give way),(do not change lanes, give way),(do not change lanes, do not give way)}.

[0026] In the lane-changing decision of autonomous vehicles, "utility" is a key concept used to quantify and evaluate the gains or losses of the vehicle under a specific strategy. The utility function is an important tool in the decision-making process, helping the autonomous driving system make the optimal lane-changing decision.

[0027] In S3, for , the utility function of the decision-making subject includes: The safety utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: (1); (2); In the formula, , are the safety utilities of the automated lane-changing vehicle and the target vehicle, respectively; d is the distance between the automated lane-changing vehicle and the target vehicle; , are the speeds of the autonomous driving lane-changing vehicle and the target vehicle at time t, respectively; , Represent the logical operators "and" and "or" respectively; The efficiency utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: (3); (4); In the formula, , are the efficiency utilities of the autonomous lane-changing vehicle and the target vehicle, respectively; The expected speed of the lane-changing vehicle and the target vehicle (the expected speeds of both are the same); , are the accelerations of the autonomous driving lane-changing vehicle and the target vehicle at time t, respectively; The comfort utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: (5); (6); In the formula, , are the comfort utilities of the autonomous lane-changing vehicle and the target vehicle, respectively; , are the accelerations of the automated lane-changing vehicle and the target vehicle at time t+4f, respectively, and f is the duration of each frame; , are the maximum acceleration and maximum deceleration of the autonomous lane-changing vehicle and the target vehicle, respectively; The fatigue utility function of the manually driven target vehicle is: (7); In the formula, Fatigue effect of manually driven target vehicle; , are the accelerations of the manually driven target vehicle at time t and time t+4f, respectively; The energy-saving utility functions of the autonomous driving lane-changing vehicle and the autonomous driving target vehicle are: (8); (9); In the formula, , They are the energy-saving utility of the autonomous lane-changing vehicle and the autonomous target vehicle, respectively; and are the frontal area and mass of the autonomous lane-changing vehicle, respectively; and are the frontal area and mass of the autonomous driving target vehicle respectively; is the quality factor; is the road slope; is the air density; is the wind speed; g is the acceleration due to gravity; , , They are rolling resistance coefficient, wind resistance coefficient and internal friction resistance coefficient; Since the dimensions of safety utility, efficiency utility, comfort utility, fatigue utility and energy saving utility are different and cannot be directly added, each utility function is normalized and its normalized value is expressed as: (10); In the formula, is the kth utility value obtained by decision subject N using G when the strategy combination is {i, j}; N={ALV, TV}, ALV and TV are the autonomous lane-changing vehicle and the target vehicle respectively; G={S, E, C, F, ES}, S, E, C, F and ES represent safety utility, efficiency utility, comfort utility, fatigue utility and energy-saving utility respectively; k={1, 2, 3, …, K}, K is the sample size of the utility value obtained by the decision subject; make , , Represent the total utility of the autonomous driving lane-changing vehicle, the total utility of the manually driven target vehicle, and the total utility of the autonomous driving target vehicle, respectively. Then: (11); (12); (13).

[0028] The utility matrix of the decision-making subject specifically includes: As shown in Table 1, the utility matrix of the autonomous lane-changing vehicle and the manually driven target vehicle is: when the autonomous lane-changing vehicle chooses to change lanes and the manually driven target vehicle chooses to give way, the utility is ; When the autonomous lane-changing vehicle chooses to change lanes and the manually driven target vehicle chooses not to give way, the utility is ; When the autonomous lane-changing vehicle chooses not to change lanes and the manually driven target vehicle chooses to give way, the utility is ; When the autonomous lane-changing vehicle chooses not to change lanes and the manually driven target vehicle chooses not to give way, the utility is .

[0029] Table 1 Utility matrix of the autonomous lane-changing vehicle and the manually driven target vehicle

[0030] That is, when both i and j are 1 , the rest are similar; As shown in Table 2, the utility matrix of the autonomous driving lane-changing vehicle and the autonomous driving target vehicle is: when the autonomous driving lane-changing vehicle chooses to change lanes and the autonomous driving target vehicle chooses to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses to change lanes and the autonomous driving target vehicle chooses not to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses not to change lanes and the autonomous driving target vehicle chooses to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses not to change lanes and the autonomous driving target vehicle chooses not to give way, the utility is .

[0031] Table 2 Utility matrix of autonomous driving lane-changing vehicle and autonomous driving target vehicle

[0032] The proportion of the lane-changing strategy and the non-lane-changing strategy selected by the autonomous driving lane-changing vehicle are recorded as p and 1-p respectively, and the proportion of the target vehicle selecting the yielding strategy and the non-yielding strategy are recorded as q and 1-q respectively.

[0033] In S4, the replication dynamic equations of the decision-making subject include: Expected utility of lane-changing strategy and non-lane-changing strategy for autonomous driving lane-changing vehicles , They are: (14); (15); Average expected utility of autonomous lane-changing vehicles choosing lane-changing strategies and non-lane-changing strategies for: (16); Replicated dynamic equations for lane-changing strategy selection by autonomous driving vehicle for: (17); Expected utility of manually driven target vehicles choosing the give way strategy and the no-give way strategy , They are: (18); (19); Average expected utility of manually driven target vehicles choosing the give way strategy and the no-give way strategy for: (20); Replicated dynamic equations for the yield strategy of manually driven target vehicles It is expressed as: (twenty one); Expected utility of autonomous driving target vehicle choosing yielding strategy and not yielding strategy , They are: (twenty two); (twenty three); Average expected utility of the autonomous driving target vehicle choosing the give way strategy and the no-give way strategy for: (twenty four); Replicated dynamic equations for autonomous driving target vehicle's choice of yield strategy for: (25).

[0034] For the replication dynamic equation of the decision-making subject, let When choosing a yielding strategy or a non-yielding strategy for the target vehicle, the utility of the lane changing strategy and the non-lane changing strategy for the autonomous driving lane changing vehicle is different, that is, in scenarios 1 and 2, , ; make When choosing a lane-changing strategy or a non-lane-changing strategy for an autonomous vehicle, the utility of the target vehicle choosing a yielding strategy or a non-yielding strategy is different, that is, in scenario 1, , In scene 2, , ; Then equation (17) is simplified to equation (26), and equations (21) and (25) are simplified to equation (27): (26); (27); In the formula, represent or .

[0035] In S5, the solution of the replicated dynamic equation is the optimization result of the evolutionary game model, based on which the specific lane-changing decision is made, including the equilibrium point and the stability of the equilibrium point, where: For the equilibrium point, the differential equations about p and q composed of equations (26) and (27) are solved to obtain equation (28), which gives the five equilibrium points of the game between the autonomous driving lane-changing vehicle and the target vehicle, namely (0,0), (0,1), (1,0), (1,1) and : (28); The Jacobian matrix J in the game is: (29); For the stability of equilibrium points, whether the equilibrium point is a stable point, a saddle point, a central point or an unstable point depends on the determinant and the sign of the trace of the Jacobian matrix; the criteria for determining the type of equilibrium point are obtained according to the Friedman method, as shown in Table 3. When the determinant of the Jacobian matrix is ​​greater than 0 and the trace is less than 0, the equilibrium point is a stable point, and the strategy selected by the decision-making subject is an evolutionary stable strategy.

[0036] Table 3 Criteria for determining the type of equilibrium point in evolutionary games

[0037] Table 4 gives the determinant and trace of the Jacobian matrix of the evolutionary game of the lane-changing vehicle of the autonomous driving. It can be seen that the positive and negative of the determinant and trace of the Jacobian matrix of the evolutionary game of the lane-changing vehicle of the autonomous driving is determined by , , , positive and negative.

[0038] Table 4 Evolutionary game equilibrium points and Jacobian matrix determinants and traces

[0039] As shown in Table 5, based on the utility function and the utility matrix, in scenario 1, The eight situations are as follows: Scenario 1: , , and , the stable point is (1,1); Scenario 2: , , and , the stable points are (0,0) and (1,1); Scenario 3: , , and , the stable point is (0,1); Scenario 4: , , and , the stable point is (0,0); Scenario 5: , , and , no stable point; Scenario 6: , , and , the stable point is (0,0); Scenario 7: , , and , the stable point is (0,1); Scenario 8: , , and , the stable point is (0,0); In scenario 2, and Always true, meets the conditions 1 to 4.

[0040] Table 5 Stability analysis of the equilibrium point of evolutionary game

[0041] Assuming the initial values ​​of p and q to be 0~1 respectively, the game evolution results are obtained by numerical simulation; according to the stability analysis of the game equilibrium point, the evolutionary stable strategies of the game system include the following three combinations: (lane change, give way), (no lane change, no give way) and (no lane change, give way); Since the strategy combination (lane change, no giving way) will bring safety risks to both decision-makers, the corresponding equilibrium point cannot be a stable point; The autonomous lane-changing vehicle makes behavioral decisions based on the strategy corresponding to the evolutionary game results: first, identify the evolutionarily stable strategy; then, determine whether the currently selected strategy is an evolutionarily stable strategy based on the current operating status. If the currently selected strategy is an evolutionarily stable strategy, then maintain the currently selected strategy; if the currently selected strategy is not an evolutionarily stable strategy, then select the evolutionarily stable strategy; if there are multiple evolutionarily stable strategies, then select the strategy with the largest total utility.

[0042] The verification process is to use a DJI Mavic 2 Zoom drone to shoot the traffic flow in a certain expressway merging area, record the vehicle entry and exit conditions and driving trajectories, as well as road traffic environment information. Figure 4 Shows the scene shot by drone. Figure 4 In the scenario, the main road is a six-lane bidirectional road. The vehicle in the current lane (ramp) is an autonomous lane-changing vehicle. The rightmost lane of the main road is the target lane. The lane-changing decision strategy for the autonomous lane-changing vehicle in this scenario needs to be completed. The Tracker software is used to extract data from the video to obtain the driving speed, acceleration, and position coordinates (x, y) of the autonomous lane-changing vehicle and the target vehicle, and store them in the form of a database.

[0043] From the database, extract data that matches the game between the autonomous driving lane-changing vehicle and the target vehicle. If the autonomous driving lane-changing vehicle does not occupy the target lane and the target vehicle reaches the conflict zone, there is a game between the two, and the judgment conditions are: (30); (31); In the formula, The longitudinal coordinate of the vehicle for autonomous lane change; is the longitudinal coordinate of the centerline of the target lane.

[0044] Among them, the three-second rule is used to determine whether the target vehicle has reached the conflict zone, such as Figure 5 shown. Figure 5 In the figure, it is assumed that the autonomous lane-changing vehicle is at point A and the target vehicle is at point B at time t. If the target vehicle reaches point A within 3 seconds, the target vehicle needs to brake and slow down to avoid colliding with the autonomous lane-changing vehicle. That is, the lane-changing behavior of the autonomous lane-changing vehicle affects the driving of the target vehicle. At this time, the target vehicle reaches the conflict zone.

[0045] The acquired data is used to calculate the utility of the automated lane-changing vehicle and the target vehicle according to equations (1) to (13), and the statistical results of different types of utility are obtained, as shown in Table 6.

[0046] Table 6 Statistical results of different types of utility

[0047] In Table 6, the dimensions of the average, maximum and minimum values ​​of safety utility and efficiency utility are all seconds; the dimensions of the average, maximum and minimum values ​​of fatigue utility are all meters per second cubed; the dimensions of the average, maximum and minimum values ​​of energy-saving utility are all kilowatts per kilogram.

[0048] The utility matrix of the decision-making subject is calculated. Table 7 shows the utility matrix of the autonomous lane-changing vehicle and the manually driven target vehicle, and Table 8 shows the utility matrix of the autonomous lane-changing vehicle and the autonomous driving target vehicle.

[0049] Table 7 Utility matrix of autonomous lane-changing vehicles and manually driven target vehicles

[0050] Table 8 Utility matrix of autonomous driving lane-changing vehicle and autonomous driving target vehicle

[0051] The replicated dynamic equations of the autonomous lane-changing vehicle, the manually driven target vehicle, and the autonomous driving target vehicle are: (32); (33); (34); The evolutionary game model is solved, and specific lane-changing decisions are made based on the results to ensure that the lane-changing goal is effectively achieved under the premise of safety. The five equilibrium points of the game between the autonomous driving vehicle and the target vehicle can be obtained from Equation (28). In scenario 1, the five equilibrium points are (0,0), (0,1), (1,0), (1,1), and (0.8099, 0.8744); in scenario 2, the five equilibrium points are (0,0), (0,1), (1,0), (1,1), and (0.7166, 0.8744).

[0052] The Jacobian matrices J1 and J2 of the game systems in scenario 1 and scenario 2 are: (35); (36); Table 9 lists the determinants and traces of the five equilibrium points corresponding to scenario 1 and scenario 2. According to the determination of the equilibrium point type in Table 3, it is concluded that the stable point of the game system is (0,0) or (1,1). Figure 6 The evolution path of the game between the autonomous driving lane-changing vehicle and the target vehicle is shown. Scenario 1 and Scenario 2 are consistent with Figure 6 This is consistent with the second case in Table 5.

[0053] Table 9 Determinant and trace of equilibrium point

[0054] The autonomous lane-changing vehicle makes behavioral decisions based on the strategy corresponding to the evolutionary game results. The vehicle's current operating state meets , , , if the game system is at the equilibrium point (0,0) or (1,1), that is, the decision-making subject chooses the strategy combination (not changing lanes, not giving way) or (changing lanes, giving way), and the total utility of the two strategy combinations is compared, the total utility of the strategy combination (not changing lanes, not giving way) is greater than the strategy combination (changing lanes, giving way), so the autonomous driving vehicle chooses not to change lanes. If the game system is not at the equilibrium point (0,0) or (1,1), then the autonomous driving vehicle adjusts the currently selected strategy to not changing lanes. The game system guides the decision-making subject's strategy selection to gradually evolve to the strategy combination (not changing lanes, not giving way).

[0055] Figure 7 The evolution paths of lane changing for autonomous vehicles corresponding to scenarios 1 and 2 are shown. Figure 7 As shown in the figure, the evolution path of the dynamic game shows obvious stage characteristics. When p and q are small, the game system is in an unstable state, p and q change greatly, the automatic driving lane-changing vehicle and the target vehicle are constantly testing and adjusting strategies. During the evolution process, the proportion of lane-changing strategies or giving way strategies will gradually decrease, and finally converge to the strategy combination (no lane change, no giving way); when p and q are large, the game system gradually tends to a stable state, and the proportion of lane-changing strategies or giving way strategies will gradually increase, and finally converge to the strategy combination (lane change, giving way).

[0056] Scenario 2 converges to a stable state faster than Scenario 1, indicating that the interactive decision-making efficiency between the autonomous lane-changing vehicle and the target vehicle is higher. At the same time, the increase in the selection frequency of the strategy combination (lane change, give way) further shows that in the game between autonomous vehicles, both parties are more likely to choose the strategy combination (lane change, give way), which reflects that when interactive decision-making between autonomous vehicles, both parties are more inclined to choose a cooperative strategy in order to achieve the best overall utility.

[0057] Whether it is scenario 1 or scenario 2, the evolution path reflects the process of the vehicle continuously adjusting its strategy based on the utility function and the replication dynamic equation. The final selected strategy combination is consistent with the evolutionary stable strategy analyzed in Table 5, which further verifies the rationality and stability of the model in this embodiment.

Claims

1. An evolutionary game decision-making method for lane changing of an autonomous driving vehicle, characterized in that The following steps are involved: S1. When the autonomous driving lane-changing vehicle changes lanes from the current lane to the target lane, predict the target vehicle of the autonomous driving lane-changing vehicle, collect the motion information of the autonomous driving lane-changing vehicle and the target vehicle in real time through sensors, and determine the game scenario of the autonomous driving lane-changing vehicle changing lanes; S2. Define the autonomous lane-changing vehicle and the target vehicle as decision-making entities, determine the strategy set of the autonomous lane-changing vehicle and the target vehicle, and the strategy combination of the game between the autonomous lane-changing vehicle and the target vehicle; S3. Construct an evolutionary game model for interactive decision-making between the autonomous lane-changing vehicle and the target vehicle, including the utility function and utility matrix of the autonomous lane-changing vehicle and the target vehicle; S4. Establish a replica dynamic equation of the decision-making subject, simulate the strategy selection process of the decision-making subject, and adjust its strategy selection according to the utility of each strategy; S5. Make and execute lane-changing decisions based on the optimization results obtained from the evolutionary game model.

2. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 1, characterized in that: In the above-mentioned S1, the motion information includes position, speed, and acceleration.

3. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 1, characterized in that: In the above S1, the game scenario is specifically as follows: On a one-way two-lane road or above, if the utility obtained by the vehicle in the current lane is lower than the utility obtained in the target lane, the vehicle will make a lane-changing decision analysis based on the operating status of itself and surrounding vehicles to decide whether to change lanes to the target lane. At this time, the vehicle in the current lane is the lane-changing vehicle, and the rear vehicle in the target lane is the target vehicle. If the behavior decisions of the lane-changing vehicle and the target vehicle can be accurately predicted, no game occurs between the lane-changing vehicle and the target vehicle. Specifically, if the lane-changing vehicle has started to change lanes and occupies the target lane, the target vehicle can only give way while the lane-changing vehicle continues to complete the lane change. At this time, no game occurs between the lane-changing vehicle and the target vehicle. If the target vehicle has not reached the conflict zone, the lane-changing vehicle performs the lane-changing operation, and the target vehicle does not need to brake or slow down and drives normally. At this time, there is no game between the lane-changing vehicle and the target vehicle. Except for these two cases, there is a game between the lane-changing vehicle and the target vehicle. Considering that the lane-changing vehicle is an autonomous driving vehicle, and the target vehicle is a manually driven vehicle or an autonomous driving vehicle, the game scenarios of autonomous driving vehicle lane changing are divided into scenario 1 and scenario 2. Scenario 1: the game between the autonomous driving lane-changing vehicle and the manually driven target vehicle, and scenario 2: the game between the autonomous driving lane-changing vehicle and the autonomous driving target vehicle.

4. The evolutionary game decision method for lane changing of an autonomous driving vehicle according to claim 3, characterized in that: In S2, the strategy set for the automatic lane-changing vehicle is {A i |i=1,2}={lane change, no lane change}, the target vehicle’s strategy set is {T j |j=1,2}={give way, don’t give way}; the strategy combination of the game between the autonomous driving lane-changing vehicle and the target vehicle is {(A i ,T j )|i,j∈{1,2}}={(change lanes, give way),(change lanes, do not give way),(do not change lanes, give way),(do not change lanes, do not give way)}.

5. The evolutionary game decision method for lane changing of an autonomous driving vehicle according to claim 4, characterized in that: In S3, for , the utility function of the decision-making subject includes: The safety utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: （1）； （2）； In the formula, , are the safety utilities of the automated lane-changing vehicle and the target vehicle, respectively; d is the distance between the automated lane-changing vehicle and the target vehicle; , are the speeds of the autonomous driving lane-changing vehicle and the target vehicle at time t, respectively; , Represent the logical operators "and" and "or" respectively; The efficiency utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: （3）； （4）； In the formula, , are the efficiency utilities of the autonomous lane-changing vehicle and the target vehicle, respectively; The desired speeds of the lane-changing vehicle and the target vehicle for automated driving; , are the accelerations of the autonomous driving lane-changing vehicle and the target vehicle at time t, respectively; The comfort utility functions of the autonomous driving lane-changing vehicle and the target vehicle are: （5）； （6）； In the formula, , are the comfort utilities of the autonomous lane-changing vehicle and the target vehicle, respectively; , are the accelerations of the automated lane-changing vehicle and the target vehicle at time t+4f, respectively, and f is the duration of each frame; , are the maximum acceleration and maximum deceleration of the autonomous lane-changing vehicle and the target vehicle, respectively; The fatigue utility function of the manually driven target vehicle is: （7）； In the formula, Fatigue effect of manually driven target vehicle; , are the accelerations of the manually driven target vehicle at time t and time t+4f, respectively; The energy-saving utility functions of the autonomous driving lane-changing vehicle and the autonomous driving target vehicle are: （8）； （9）； In the formula, , They are the energy-saving utility of the autonomous lane-changing vehicle and the autonomous target vehicle, respectively; and are the frontal area and mass of the autonomous lane-changing vehicle, respectively; and are the frontal area and mass of the autonomous driving target vehicle respectively; is the quality factor; is the road slope; is the air density; is the wind speed; g is the acceleration due to gravity; , , They are rolling resistance coefficient, wind resistance coefficient and internal friction resistance coefficient; Each utility function is normalized, and its normalized value is expressed as: （10）； In the formula, is the kth utility value obtained by decision subject N using G when the strategy combination is {i, j}; N={ALV, TV}, ALV and TV are the autonomous lane-changing vehicle and the target vehicle respectively; G={S, E, C, F, ES}, S, E, C, F and ES represent safety utility, efficiency utility, comfort utility, fatigue utility and energy-saving utility respectively; k={1, 2, 3, …, K}, K is the sample size of the utility value obtained by the decision subject; make , , Represent the total utility of the autonomous driving lane-changing vehicle, the total utility of the manually driven target vehicle, and the total utility of the autonomous driving target vehicle, respectively. Then: （11）； （12）； （13）。 6. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 5, characterized in that: In S3, the utility matrix of the decision-making subject specifically includes: The utility matrix of the autonomous lane-changing vehicle and the manually driven target vehicle. When the autonomous lane-changing vehicle chooses to change lanes and the manually driven target vehicle chooses to give way, the utility is ; When the autonomous lane-changing vehicle chooses to change lanes and the manually driven target vehicle chooses not to give way, the utility is ; When the autonomous lane-changing vehicle chooses not to change lanes and the manually driven target vehicle chooses to give way, the utility is ; When the autonomous lane-changing vehicle chooses not to change lanes and the manually driven target vehicle chooses not to give way, the utility is ; The utility matrix of the autonomous lane-changing vehicle and the autonomous target vehicle. When the autonomous lane-changing vehicle chooses to change lanes and the autonomous target vehicle chooses to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses to change lanes and the autonomous driving target vehicle chooses not to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses not to change lanes and the autonomous driving target vehicle chooses to give way, the utility is ; When the autonomous driving lane-changing vehicle chooses not to change lanes and the autonomous driving target vehicle chooses not to give way, the utility is ; The proportion of the lane-changing strategy and the non-lane-changing strategy selected by the autonomous driving lane-changing vehicle are recorded as p and 1-p respectively, and the proportion of the target vehicle selecting the yielding strategy and the non-yielding strategy are recorded as q and 1-q respectively.

7. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 6, characterized in that: In S4, the replication dynamic equation of the decision-making subject includes: Expected utility of lane-changing strategy and non-lane-changing strategy for autonomous driving lane-changing vehicles , They are: （14）； （15）； Average expected utility of autonomous lane-changing vehicles choosing lane-changing strategies and non-lane-changing strategies for: （16）； Replicated dynamic equations for lane-changing strategy selection by autonomous driving vehicle for: （17）； Expected utility of manually driven target vehicles choosing the give way strategy and the no-give way strategy , They are: （18）； （19）； Average expected utility of manually driven target vehicles choosing the give way strategy and the no-give way strategy for: （20）； Replicated dynamic equations for the yield strategy of manually driven target vehicles It is expressed as: （21）； Expected utility of autonomous driving target vehicle choosing yielding strategy and not yielding strategy , They are: （22）； （23）； Average expected utility of the autonomous driving target vehicle choosing the give way strategy and the no-give way strategy for: （24）； Replicated dynamic equations for autonomous driving target vehicle's choice of yield strategy for: （25）。 8. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 7, characterized in that: For the replication dynamic equation of the decision-making subject, let When choosing a yielding strategy or a non-yielding strategy for the target vehicle, the utility of the lane changing strategy and the non-lane changing strategy for the autonomous driving lane changing vehicle is different, that is, in scenarios 1 and 2, , ; make When choosing a lane-changing strategy or a non-lane-changing strategy for an autonomous vehicle, the utility of the target vehicle choosing a yielding strategy or a non-yielding strategy is different, that is, in scenario 1, , In scene 2, , ; Then equation (17) is simplified to equation (26), and equations (21) and (25) are simplified to equation (27): （26）； （27）； In the formula, represent or .

9. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 8, characterized in that: In S5, the solution of the replicated dynamic equation is the optimization result of the evolutionary game model, and a specific lane-changing decision is made based on it, including the equilibrium point and the stability of the equilibrium point, where: For the equilibrium point, the differential equations about p and q composed of equations (26) and (27) are solved to obtain equation (28), which gives the five equilibrium points of the game between the autonomous driving lane-changing vehicle and the target vehicle, namely (0,0), (0,1), (1,0), (1,1) and : （28）； The Jacobian matrix J in the game is: （29）； Regarding the stability of the equilibrium point, whether the equilibrium point is a stable point, a saddle point, a central point or an unstable point depends on the determinant and the sign of the trace of the Jacobian matrix; the judgment criterion of the equilibrium point type is obtained according to the Friedman method. When the determinant of the Jacobian matrix is ​​greater than 0 and the trace is less than 0, the equilibrium point is a stable point, and the strategy selected by the decision-making subject is an evolutionary stable strategy.

10. The evolutionary game decision-making method for lane changing of an autonomous driving vehicle according to claim 9, characterized in that: The Jacobian matrix determinant and trace of the evolutionary game of lane-changing vehicles in autonomous driving are positive or negative depending on , , , The positive and negative; Based on the utility function and utility matrix, in scenario 1, The eight situations are as follows: Scenario 1: , , and , the stable point is (1,1); Scenario 2: , , and , the stable points are (0,0) and (1,1); Scenario 3: , , and , the stable point is (0,1); Scenario 4: , , and , the stable point is (0,0); Scenario 5: , , and , no stable point; Scenario 6: , , and , the stable point is (0,0); Scenario 7: , , and , the stable point is (0,1); Scenario 8: , , and , the stable point is (0,0); In scenario 2, and It is always established and meets the conditions 1 to 4; Assuming the initial values ​​of p and q to be 0~1 respectively, the game evolution results are obtained by numerical simulation; according to the stability analysis of the game equilibrium point, the evolutionary stable strategies of the game system include the following three combinations: (lane change, give way), (no lane change, no give way) and (no lane change, give way); Since the strategy combination (lane change, no giving way) will bring safety risks to both decision-makers, the corresponding equilibrium point cannot be a stable point; The autonomous lane-changing vehicle makes behavioral decisions based on the strategy corresponding to the evolutionary game results: first, it identifies the evolutionarily stable strategy; then, based on the current operating state, it determines whether the currently selected strategy is an evolutionarily stable strategy. If the currently selected strategy is an evolutionarily stable strategy, then it maintains the currently selected strategy; If the currently selected strategy is not an evolutionary stable strategy, then select an evolutionary stable strategy; if there are multiple evolutionary stable strategies, then select the strategy with the largest total utility.

Images