A lane-changing decision-making method based on game theory considering different vehicle types

Abstract

The present application belongs to the technical field of intelligent vehicle decision-making, and relates to a lane-changing decision-making method based on game theory, which takes different vehicle types into consideration. In the method, vehicle information is collected by a vehicle-mounted data collection device of the intelligent vehicle; lane-changing decision-making is determined based on the collected information; a lane-changing request is informed through V2V communication; game participants are determined, and a driving intention set is established for both parties of the game; driving benefits in the lane-changing process are determined according to different vehicle types; a game between the current vehicle and a vehicle behind the target lane is performed, and a lane-changing decision-making model is established; the benefits corresponding to different driving intentions of the lane-changing vehicle and the lagging vehicle are calculated according to the benefit function defined above; Nash equilibrium is solved, and the final driving decision is obtained. The present application helps to improve road traffic efficiency, safety and feasibility of intelligent vehicle decision-making results.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent vehicle decision-making, and specifically refers to a lane-changing decision-making method based on game theory that takes into account different vehicle models. Background Art

[0002] Intelligent vehicle decision-making is a key technology for autonomous driving. This involves determining the optimal driving intention and route based on information about the surrounding environment. During this process, the intelligent vehicle interacts with its surrounding vehicles, so game theory is used to guide its decision-making. Based on this game theory approach, the vehicle is considered a player in the game. The strategies of the lane-changing vehicle and surrounding vehicles are explored, and the lane-changing vehicle and its surrounding vehicles engage in a non-cooperative game under perfect information conditions. For the lane-changing problem, we incorporate game theory into the vehicle's lane-changing strategy, taking into account the interactions between the vehicle and surrounding vehicles. The vehicle selects its strategy based on self-interest. Given a set of vehicle strategies, the vehicle can obtain real-time information about surrounding road conditions, other vehicles' characteristics, and possible actions through onboard data acquisition equipment. By solving the Nash equilibrium of the intelligent vehicle decision-making problem, we can determine the most likely behavior of surrounding vehicles and the decision-making behavior of the host vehicle. The application of Nash equilibrium assumes that all players act in their own self-interest and choose the optimal strategy for themselves, i.e., adopting a self-optimal decision-making model. However, in practice, different types of vehicles have different motivations and decision preferences for lane changes. Small vehicles prioritize speed, while large vehicles prioritize stability. Therefore, it's impossible to establish a single payoff function for all vehicles in the game. Therefore, designing different payoff functions based on vehicle type to guide lane-changing decisions can help improve the safety and feasibility of intelligent vehicle decision-making. Summary of the Invention

[0003] To solve the above problems, the present invention provides a lane-changing decision method based on game theory that takes into account the differences in specific performance and driving behavior between small vehicles and large vehicles.

[0004] The technical solution of the present invention:

[0005] A lane-changing decision-making method based on game theory considering different vehicle types is proposed. The specific steps are as follows:

[0006] Step (1) collects the vehicle information of the smart car and its surrounding vehicles through the vehicle-mounted data acquisition equipment (such as radar, camera and lidar).

[0007] The road scene considered in this invention is a one-way two-lane road that is straight and contains different types of vehicles. i , i=1,2,…,N, the information set can be described as:

[0008] χ (t) = {x i (t),v i (t),a i (t),y i iρ i}

[0009] x i (t), v i (t) and a i (t) represents the vehicle V i The longitudinal position, longitudinal velocity and longitudinal acceleration at time t define the longitudinal position x of the starting end of the current road i (t) = 0, where longitudinal refers to the direction of vehicle travel, y i Indicates vehicle V i Lane, ρ i Indicates vehicle V i The type of vehicle is defined as Type I and Type II according to the vehicle length. Among them, vehicles with a length of less than 4.3 meters are Type I, and vehicles with a length of more than 5.9 meters are Type II.

[0010] Step (2) makes a lane-changing decision based on the collected information.

[0011] Vehicle V i According to the surrounding traffic conditions, including the speed and position information of surrounding vehicles, it is determined whether the current lane meets the driving requirements of the vehicle. If not, the vehicle V i The initial lane-changing motivation will be generated, and the road conditions of the adjacent lanes will be further judged to see whether they support the lane-changing behavior. If so, V i The final lane-changing intention; on the contrary, V i You can only continue driving in the original lane and do not need to make subsequent lane change decisions. The specific judgment method is as follows:

[0012] Vehicle I is more interested in speed gains during driving, that is, it hopes to travel at a faster speed. When the vehicle ahead affects the acceleration of Vehicle I, Vehicle I will be motivated to change lanes, which means that the following conditions are met:

[0013]

[0014] where x i (t) is the vehicle V at time t i location, V at time t i The position of the vehicle directly ahead, d ef Indicates the maximum distance that is not affected when driving. Beyond this distance, the vehicle ahead will affect the driving decision of the vehicle. i (t) is the vehicle V at time t iThe speed of the vehicle in the driving direction, Vt i The speed of the vehicle in the driving direction.

[0015] Unlike vehicle I, vehicle II pursues not only the speed gain but also the stability of its own driving, avoiding frequent changes in its own conditions due to frequent lane changing. Therefore, vehicle II has a higher tolerance for the front vehicle, and only when the front vehicle affects the current speed driving of vehicle II, vehicle II will generate a lane changing motive, i.e. to meet the following conditions:

[0016]

[0017] After meeting the lane changing motive, it is also necessary to judge whether the road condition of the adjacent lane meets the lane changing requirement. The adjacent lane of the current lane is referred to as the target lane, and the first vehicle behind the target lane is referred to as the lagging vehicle. The road condition mainly refers to whether the distance between the current vehicle and the lagging vehicle meets the safety condition, so as to avoid rear-end collision accidents with the lagging vehicle on the target lane during the lane changing process of the current vehicle. The safety distance needs to be guaranteed in the worst case, i.e. the vehicle V i adopts emergency braking, and the lagging vehicle still will not be rear-ended. Taking vehicle I as an example, at this time, the following conditions need to be met:

[0018]

[0019] wherein Vt i is the position of the vehicle head of V i at time t, x i is the position of the vehicle head of V i at time t, and l i is the length of the vehicle body of V i . Let dis represent the distance between the two vehicles at time t, and further obtain:

[0020]

[0021] wherein dis i (t) and dis i (t) represent the emergency braking distance of V i,safe and V i,back at time t, respectively, and b is the emergency braking acceleration of the vehicle. Therefore, the safety driving condition of V i,safe is that the distance between the two vehicles is not less than the difference between the braking distance of the lagging vehicle and the braking distance of the vehicle. Considering that vehicle II is more important than vehicle I, and has poor emergency braking capability, a larger space should be required as the safety distance. The right side of formula (1) is referred to as dis i (t), and the final safety constraint condition is obtained:

[0022] Δχ i,back (t)≥η·dis i,safe (t)

[0023] Here, η is called the safety factor. The larger the safety factor, the higher the vehicle's safety requirements. The η of vehicle type II is greater than that of vehicle type I.

[0024] After the lane-changing motivation and safety conditions are met, the vehicle conducts V2V communication to inform surrounding vehicles of the lane-changing request.

[0025] Step (3) determines the game participants and establishes a strategy combination for both parties.

[0026] Game theory is used in lane-changing decisions. i It is necessary to play a game with the lagging vehicle in the target lane to improve the lane change success rate and reduce the occurrence of congestion. Therefore, the game participants are determined to be the lane-changing vehicle (EV) and the lagging vehicle (SV) in the target lane. The game between the two parties can be expressed as a four-tuple Γ = <P,A P ,S P ,E P >. Where P is the set of game participants, P = {lane-changing vehicle (EV), lagging vehicle (SV)}, where the types of EV and SV are model I and model II. P Refers to the set of actions available to game participants. The driving action set of EV is A EV = {LC, DC}, LC means lane change, DC means no lane change. SV can choose to accept or reject the lane change behavior of EV. If it chooses to accept, it will be realized by slowing down or driving at a constant speed; if it chooses to reject, it will accelerate. Therefore, the driving action set of SV is A SV ={UN, AC, DE}, UN represents uniform speed, AC represents acceleration, and DE represents deceleration. P It is a set of strategy combinations of both parties in the game. There are six strategy combinations between the two parties in the game, so S P ={(no lane change, acceleration); (no lane change, deceleration); (no lane change, constant speed); (lane change, acceleration); (lane change, deceleration); (lane change, constant speed)}. E P It is the set of payoffs of the game participants.

[0027] Step (4) Analyze the benefits of both parties in the game.

[0028] EV and SV achieve different returns when adopting different pure strategies and Where i = {1, 2}, j = {1, 2, 3}. t = {C, T} indicates that the vehicle types are Model I and Model II. The specific expression of the profit matrix is

[0029] Furthermore, different vehicle types produce different payoffs. In the lane-changing game between EVs and SVs, each player knows the other's possible action sets. The EV and SV will choose different actions with certain probability distributions based on the circumstances. Furthermore, the two players have different driving goals and aspirations, each seeking to maximize their own payoffs and choose the strategy that best benefits them. Therefore, the game is a complete information, non-cooperative, mixed-strategy game.

[0030] In a mixed strategy game, EV and SV randomly select different strategies with a certain probability, where p1 and p2 represent the probabilities of EV adopting the pure strategies of changing lanes and not changing lanes, respectively, and p2 = 1-p1; q1, q2, and q3 represent the probabilities of SV adopting the pure strategies of accelerating, decelerating, and maintaining a constant speed, respectively, and q3 = 1-q1-q2. Based on the payoffs of the two players adopting different strategies in the game and the probabilities of EV changing lanes and SV accelerating and decelerating, the payoff matrix for the mixed strategy case is expressed as follows:

[0031] The expected return of the hybrid strategy of vehicle EV and vehicle SV is E EV 、E SV It can be expressed as follows:

[0032]

[0033] Step (5) designs a profit function based on the factors that affect the vehicle's lane-changing behavior.

[0034] The factors that affect the vehicle's lane-changing decision mainly include the speed difference between the vehicle and the front and rear vehicles, the distance between the vehicle and the front and rear vehicles, etc. Therefore, when formulating the benefit function, the speed benefit U velocity and space benefit U space Set as the main decision factor of the target vehicle. In addition, vehicle type I pays attention to the type of vehicle in front and is unwilling to drive behind vehicle type II, so the benefit of the vehicle type in front is considered. Model II focuses on the stability of its own driving state, including speed stability and dimensional stability Different types of vehicles have different driving characteristics, different impacts on the road, and different corresponding benefit functions. The benefit of the current vehicle in step (4) is and the payoff of the lagging vehicle is The cases of t = {C, T} are discussed separately. The benefits obtained when the current vehicle (EV) is model I and model II and the target lane following vehicle (SV) is model I and model II are analyzed.

[0035] First, consider the case where the EV is model I and chooses to change lanes, mainly considering the speed gain. and space benefits Since vehicle type I is affected by the type of vehicle in front, the benefit of the vehicle type in front must also be considered. If the lagging vehicle decides to accelerate or decelerate, the safety condition Δχ determined in step (2) i,back (t)≥η·dis i,safe (t) changes, so security benefits need to be considered The benefits generated by the above conditions are specifically expressed as follows:

[0036]

[0037]

[0038]

[0039] in ξ t Qualitatively represents the type of vehicle ahead after lane change, ξ t-1 Indicates the type of vehicle before lane change; ξ for vehicle type I is positive, ξ for vehicle type II is negative, and when EV performs lane change The benefit of lane changing is evaluated by comparing the types of vehicles before and after the lane change. The speed benefit is specifically expressed as where v front (t) represents the speed of the preceding vehicle at time t after lane change, v(t) represents the speed of the vehicle at time t, and v max Represents the road speed limit. The vehicle expects to achieve a higher driving speed. If the speed of the preceding vehicle is lower than its own speed, the benefit is negative, otherwise it is positive. The spatial benefit is specifically expressed as where x front (t) represents the front position of the preceding vehicle at time t, x(t) represents the front position of the vehicle at time t, l front The greater the distance between the vehicle and the preceding vehicle, the more operating space and reaction time the driver has, and the higher the driving safety and comfort. where v i,back (t) is the speed of the following vehicle after the lane change. Each benefit U is normalized to satisfy -1≤U≤1. are the parameters that need to be adjusted.

[0040] Consider the case where the EV is a vehicle type II and chooses to change lanes. Except for not considering the benefit of the vehicle type in front, the other conditions are the same as those for the vehicle type I lane change, specifically expressed as:

[0041]

[0042]

[0043]

[0044] are the parameters that need to be adjusted.

[0045] When the EV is Model I and chooses not to change lanes, the speed gain is still considered. and space benefits At this time, the type of vehicle in front of the EV does not change, so Since the EV does not change lanes, the vehicle behind Model I remains unchanged. At this time, the benefit of Model I does not consider the impact of the vehicle behind it.

[0046] When the EV is a Type II vehicle and chooses not to change lanes, in addition to considering the speed gain and space benefits In addition, spatial stability benefits must also be considered Where x back (t) is the longitudinal position of the front of the following vehicle at time t, l is the length of the vehicle body, d safe Indicates safe distance; the larger the space behind the vehicle, the smaller the chance of a rear-end collision. The greater this benefit, the better the space stability.

[0047] Consider the case where the SV is vehicle type I and it chooses to accelerate. Vehicle type I is affected by the type of vehicle in front. If the EV changes lanes, the type of vehicle in front may change. When EV does not change lanes, it is also necessary to consider the benefits of the vehicle type in front. The benefits generated by the above conditions are specifically expressed as follows:

[0048]

[0049]

[0050] Furthermore, the benefit of the SV being vehicle type I and choosing to decelerate is the same as in the above case.

[0051] When SV is vehicle type II and acceleration is considered, vehicle type II expects its driving state to be stable, and the benefit is affected by speed changes. Speed ​​stability is defined as Where v(t-1) represents the speed before SV deceleration, v max Indicates the vehicle speed limit. Excessive speed changes will affect the smooth driving of Model II. The benefits generated by the above conditions are specifically expressed as:

[0052]

[0053]

[0054] In addition, the benefit when SV is vehicle type I and chooses to decelerate is the same as the above case; when SV is at a constant speed, the speed does not change, so it is not considered.

[0055] Step (6) solves the Nash equilibrium and obtains the final driving decision.

[0056] For mixed strategies, it is impossible for a participant to obtain a better payoff than the Nash equilibrium by changing only his own mixed strategy without changing the other party's strategy. Therefore, the Nash equilibrium can be expressed as follows:

[0057]

[0058]

[0059] at this time and is the expected return of EV and SV at Nash equilibrium. According to the return matrix in step (4), the returns of EV and SV can be expressed as:

[0060]

[0061]

[0062] Among them E EV (p1,q1,q2) is the expected return of EV, E SV (p1,q1,q2) is the expected return of SV.

[0063] According to the expected returns of EV and SV, the Nash equilibrium can be further expressed as:

[0064]

[0065]

[0066] After solving the Nash equilibrium, we get (p1 * ,q1 * ,q2 * ) probability of the strategy, p1 * is the best response of EV to SV’s strategy, that is, the probability of EV changing lanes at Nash equilibrium; (q1 * ,q2 * ) is the best response of SV to EV, that is, the probability of SV choosing various actions in Nash equilibrium. Therefore, when EV takes the action with probability p1 * Lane change, 1-p1 * Without changing lanes, SV takes q1 * Acceleration, q2 * Slowdown, 1-q1 * -q2* When the speed is constant, it is the best strategy combination for the lane-changing vehicle and the vehicle behind in the target lane.

[0067] The beneficial effects of the present invention are as follows: in the decision-making process of smart cars based on game theory, the present invention takes into account the influence of driving characteristics of different models, proposes different profit functions according to different vehicle types, and solves the problem that the decision results of different types of vehicles selecting the same profit function in the traditional game theory decision-making process are inconsistent with the actual situation, which helps to improve road traffic efficiency and the safety and feasibility of smart car decision-making results. DETAILED DESCRIPTION

[0068] The specific implementation of the present invention is further described below in conjunction with the technical solution.

[0069] The present invention provides a lane-changing decision-making method based on game theory considering different vehicle models, and the specific steps are as follows:

[0070] Step (1) collects the vehicle information of the smart car and its surrounding vehicles through the vehicle-mounted data acquisition equipment of the smart car.

[0071] The vehicle-mounted data acquisition equipment includes: a positioning module, a speed sensor, a radar, a camera and a laser radar. The collected information includes the front position, speed, acceleration and vehicle type of the smart car; the front position and speed, acceleration and vehicle type of the surrounding vehicles. Positioning errors and communication delays may occur during the acquisition process, but the effects of errors and delays are ignored in the specific calculation. The road scene considered in the present invention is a one-way two-lane road with a total length of 1000m. The road is straight and contains different types of vehicles. For the current road vehicle V i , i=1,2,…,N, the information set can be described as:

[0072] χ i (t) = {x i (t),v i (t),a i (t),y i ,ρ i}

[0073] x i (t), v i (t) and a i (t) represents the vehicle V i The longitudinal position, longitudinal velocity and longitudinal acceleration at time t define the longitudinal position x of the starting end of the current road i (t) = 0, where longitudinal refers to the direction of vehicle travel, y i Indicates vehicle V i Lane, ρ i Indicates vehicle V iIn the present invention, a vehicle with a length of less than 4.3 meters is defined as a model I, such as a sedan, and a vehicle with a length of more than 5.9 meters is defined as a model II, such as a large truck.

[0074] Step (2) makes a lane-changing decision based on the collected information.

[0075] Vehicle V i Based on the surrounding traffic conditions, including the speed and position of surrounding vehicles, the system determines whether the current lane meets its driving requirements. Driving requirements include the pursuit of speed, safety, and comfort. Different vehicle types have different driving requirements: Vehicle Type I prioritizes speed gains and therefore desires faster speeds; Vehicle Type II prioritizes stability and is therefore more tolerant of vehicles ahead.

[0076] If the current lane does not meet the vehicle's driving requirements, the vehicle V i Then the initial lane-changing motivation is generated, and the driver further determines whether the road conditions in the adjacent lanes support the lane-changing behavior. If so, the driver determines V i The final lane-changing intention; on the contrary, V i You can only continue driving in the original lane and do not need to make subsequent lane change decisions. The specific judgment method is as follows:

[0077] The speed of the vehicle ahead in the target lane and the speed of the vehicle ahead in the current lane play a crucial role in generating lane change intentions. Existing literature suggests that vehicle type I (like a sedan) prioritizes speed gains during driving, meaning it desires to travel at a higher speed. When the vehicle ahead influences vehicle type I's acceleration, it will be motivated to change lanes, meaning the following conditions are met:

[0078]

[0079] where x i (t) is the vehicle V at time t i location, V at time t i The position of the vehicle directly ahead, d ef Indicates the maximum distance that is not affected when driving, which is 67m. Beyond this distance, the vehicle in front will affect the driving decision; v i (t) is the vehicle V at time t i Speed ​​in the direction of travel, V at time t i The speed of the vehicle directly ahead in the direction of travel.

[0080] Unlike Model I, Model II not only pursues speed gains but also seeks stability, avoiding frequent lane changes that could cause its own conditions to change (e.g., trucks). Therefore, Model II is more tolerant of vehicles ahead. It will only be motivated to change lanes if the vehicle ahead of it impacts its ability to maintain its current speed, i.e., if the following conditions are met:

[0081]

[0082] After satisfying the lane changing motivation, it is also necessary to determine whether the road conditions of the adjacent lanes meet the lane changing requirements. The adjacent lane of the current lane is called the target lane. i (t) The first vehicle behind is called the lagging vehicle. The road condition mentioned above mainly refers to whether the longitudinal distance between the current vehicle and the lagging vehicle meets the safety conditions, so as to avoid a rear-end collision with the lagging vehicle in the target lane during the lane change process of the current vehicle. The safety distance should be guaranteed in the worst case, that is, vehicle V i Emergency braking can prevent the lagging vehicle from rear-ending. In this case, the following conditions must be met:

[0083]

[0084] in is the position of the front of the lagging vehicle at time t, x i (t) is V i The position of the front of the car, l i For vehicle V i The length of the vehicle body. represents the distance between the two cars at time t, and further we can get:

[0085]

[0086] Among them, i (t) and They represent the vehicle V at time t respectively i and vehicle V i The emergency braking distance of the lagging vehicle is given by Determine that b is the vehicle's emergency braking acceleration, and the emergency braking acceleration of vehicle type I is 3.5 m / s 2 , the emergency braking acceleration of Model II is 2.5m / s 2 Therefore, the vehicle V i The safe driving condition is that the distance between the two vehicles is not less than the difference between the braking distance of the rear vehicle and the braking distance of the vehicle itself. Considering that the vehicle type II is heavier than the vehicle type I and has poor emergency braking capability, a larger space should be required as a safe distance. The right side of formula (1) is called dis i,safe (t), and obtain the final safety constraint:

[0087] Δx i,back (t)≥η·dis i,safe (t)

[0088] Here, η is called the safety factor. The larger the safety factor, the higher the vehicle's safety requirements. The η of model I is 0.6, and the η of model II is 0.8.

[0089] After satisfying the lane change motivation and safety requirements, the vehicles engage in V2V communication to notify the lagging vehicle of their lane change request. Because lane changes can affect the lagging vehicle in the target lane, both vehicles engage in a coordinated lane change strategy to improve the lane change success rate and reduce congestion.

[0090] Step (3) determines the game participants and establishes a strategy combination for both parties.

[0091] Vehicle V i Changing lanes may hinder the lagging vehicle in the target lane and even cause a rear-end collision, so it is not possible to change lanes directly. Game theory can be used to find the best strategy combination in interactive decision-making situations, so game theory methods are used in lane change decisions. i It is necessary to play a game with the lagging vehicle in the target lane to improve the lane change success rate and reduce the occurrence of congestion. Therefore, the game participants are determined to be the lane-changing vehicle (EV) and the lagging vehicle (SV) in the target lane. The game between the two parties can be expressed as a four-tuple Γ = <P,A P ,S P ,E P >. Where P is the set of game participants, P = {lane-changing vehicle (EV), lagging vehicle (SV)}, where the types of EV and SV are model I and model II respectively. P Refers to the set of actions available to game participants. The driving action set of EV is A EV = {LC, DC}, LC means lane change, DC means no lane change. SV can choose to accept or reject the lane change behavior of EV. If it chooses to accept, it can be realized by slowing down or driving at a constant speed; if it chooses to reject, it can accelerate. Therefore, the driving action set of SV is A SV ={UN, AC, DE}, UN represents uniform speed, AC represents acceleration, and DE represents deceleration. P It is a set of strategy combinations of both parties in the game. There are six strategy combinations between the two parties in the game, so S P ={(no lane change, acceleration); (no lane change, deceleration); (no lane change, constant speed); (lane change, acceleration); (lane change, deceleration); (lane change, constant speed)}. E P It is the set of payoffs of the game participants.

[0092] Step (4) Analyze the benefits of both parties in the game.

[0093] EV and SV achieve different returns when adopting different pure strategies and Where i = {1, 2}, j = {1, 2, 3}. C and T represent the vehicle types of Model I and Model II respectively. The specific expression of the profit matrix is:

[0094]

[0095] Furthermore, different vehicle types produce different payoffs. In the lane-changing game between EVs and SVs, each player knows the other's possible action sets. The EV and SV will choose different actions with certain probability distributions based on the circumstances. Furthermore, each player has different driving goals and aspirations, seeking to maximize their own payoffs and choosing the strategy that best benefits them. Therefore, the game is a complete information, non-cooperative, mixed-strategy game.

[0096] In a mixed strategy game, EV and SV randomly choose different strategies with a certain probability. Here, p1 and p2 represent the probabilities of EV adopting the pure strategies of changing lanes and not changing lanes, respectively, with p2 = 1 - p1; q1, q2, and q3 represent the probabilities of SV adopting the pure strategies of accelerating, decelerating, and maintaining a constant speed, respectively, with q3 = 1 - q1 - q2. Based on the payoffs of the two players adopting different strategies in the game, as well as the probabilities of EV changing lanes and SV accelerating or decelerating, the payoff matrix for the mixed strategy case is expressed as:

[0097]

[0098] The expected return of the hybrid strategy of vehicle EV and vehicle SV is E EV 、E SV It can be expressed as follows:

[0099]

[0100]

[0101] Step (5) designs a profit function based on the factors that affect the vehicle's lane-changing behavior.

[0102] Through investigation and statistical analysis of vehicle lane-changing behavior, it is found that the factors affecting the vehicle lane-changing decision mainly include the speed difference between the vehicle and the front and rear vehicles, the distance between the vehicle and the front and rear vehicles, etc. Therefore, when formulating the benefit function, the present invention takes the speed benefit U velocity and space benefit U space Set as the main decision factor of the target vehicle. In addition, vehicle type I pays attention to the type of vehicle in front and is unwilling to drive behind vehicle type II, so the benefit of the vehicle type in front is considered. Model II focuses on the stability of its own driving state, including speed stability and dimensional stability Different types of vehicles have different driving characteristics, different impacts on the road, and corresponding different benefit functions. The benefits achieved by the current vehicle (EV), which is type I, and the vehicle behind it (SV), which is type II, in the target lane, are analyzed.

[0103] The EV is traveling at a speed of 61 km / h. Before the EV lane change, the preceding vehicle is Type II, the relative distance between the two vehicles is 59 m, and the preceding vehicle's speed is 52 km / h. If the EV chooses to change lanes, the preceding vehicle is Type I after the lane change, the relative distance between the two vehicles is 63 m, and the preceding vehicle's speed is 63 km / h. The following vehicle (SV) after the lane change has a speed of 54 km / h.

[0104] First, consider the case where the EV is model I and chooses to change lanes, mainly considering the speed gain. and space benefits Since vehicle type I is affected by the type of vehicle in front, the benefit of the vehicle type in front must also be considered. If the lagging vehicle decides to accelerate or decelerate, the safety condition Δx determined in step (2) i,back (t)≥η·dis i,safe (t) changes, so security benefits need to be considered The benefits generated by the above conditions are specifically expressed as follows:

[0105]

[0106]

[0107]

[0108] in ξ t Qualitatively represents the type of vehicle ahead after lane change, ξ t-1 Indicates the type of vehicle before lane change; ξ for vehicle type I is 0.2, ξ for vehicle type II is -0.2, and when EV performs lane change The benefit of lane changing is evaluated by comparing the types of vehicles before and after the lane change. The speed benefit is specifically expressed as where v front (t) represents the speed of the preceding vehicle at time t after lane change, v(t) represents the speed of the vehicle at time t, and v max The speed limit of the road is 70km / h. The vehicle expects to achieve a higher driving speed. If the speed of the vehicle ahead is lower than its own speed, the benefit is negative, otherwise it is positive. The space benefit is specifically expressed as where xf ront (t) represents the front position of the preceding vehicle at time t, x(t) represents the front position of the vehicle at time t, l frontis the length of the preceding vehicle, which is Model I, and is 4m long. The greater the distance between the vehicle and the preceding vehicle, the more operating space and reaction time the driver has, and the higher the corresponding driving safety and comfort. The safety benefit is where v i,back (t) is the speed of the following vehicle after the lane change. Each benefit U is normalized to satisfy -1≤U≤1. It is a parameter that needs to be adjusted, and the initial parameter is set to 0.2.

[0109] When the EV is Model I and chooses not to change lanes, the speed gain is still considered. and space benefits At this time, the type of vehicle in front of the EV does not change, so Since the EV does not change lanes, the vehicle behind Model I remains unchanged. At this time, the benefit of Model I does not consider the impact of the vehicle behind. The benefit generated by the above conditions is specifically expressed as:

[0110]

[0111]

[0112]

[0113] SV travel speed is 54km / h, acceleration is 1.5m / s 2 If the EV does not change lanes, the vehicle type in front of the SV is Type I, the distance between the two vehicles is 130m, and the speed of the vehicle in front is 63km / h. If the EV changes lanes, the distance between the SV and the vehicle in front is 67m.

[0114] When SV is vehicle type II and acceleration is considered, vehicle type II expects its driving state to be stable, and the benefit is affected by speed changes. Speed ​​stability is defined as In the formula, v(t-1) represents the speed of the SV before acceleration, which is 54 km / h. Since the information collection equipment collects information every 1 second, if the speed of the SV is 59.4 km / h during acceleration or deceleration, excessive speed changes will affect the smooth driving of Model II. The benefits generated by the above conditions are specifically expressed as:

[0115]

[0116]

[0117] In addition, the benefit of SV is the same as above when it is model I and it chooses to decelerate. The speed during deceleration is 48.6 km / h. When SV is at a constant speed, the speed does not change, so it is not considered.

[0118] In addition, the benefits obtained when the current vehicle (EV) is model II and the vehicle behind (SV) in the target lane is model I are analyzed.

[0119] Considering that the EV is a vehicle type II and chooses to change lanes, except that the benefit of the vehicle type in front is not considered, the other conditions are the same as those for the small car lane change, specifically expressed as:

[0120]

[0121]

[0122]

[0123] are the parameters that need to be adjusted.

[0124] When the EV is a Type II vehicle and chooses not to change lanes, in addition to considering the speed gain and space benefits In addition, spatial stability benefits must also be considered Where x back (t) is the front position of the following vehicle at time t, l is the length of the vehicle, d safe Indicates the safe distance, which depends on the vehicle's current speed. The larger the space behind the vehicle, the smaller the chance of a rear-end collision. The greater this benefit, the better the spatial stability.

[0125] Consider the case where the SV is vehicle type I and it chooses to accelerate. Vehicle type I is affected by the type of vehicle in front. If the EV changes lanes, the type of vehicle in front may change. When EV does not change lanes, it is also necessary to consider the benefits of the vehicle type in front. The benefits generated by the above conditions are specifically expressed as follows:

[0126]

[0127]

[0128] Furthermore, the benefit of the SV being vehicle type I and choosing to decelerate is the same as in the above case.

[0129] When SV is vehicle type II and acceleration is considered, vehicle type II expects its driving state to be stable, and the benefit is affected by speed changes. Speed ​​stability is defined as Where v(t-1) represents the speed of SV before deceleration. Excessive speed changes will affect the smooth driving of Model II. The benefits generated by the above conditions are specifically expressed as:

[0130]

[0131]

[0132] In addition, the benefit when SV is vehicle type I and chooses to decelerate is the same as the above case; when SV is at a constant speed, the speed does not change, so it is not considered.

[0133] Step (6) solves the Nash equilibrium and obtains the final driving decision.

[0134] The core problem of the game is calculating the Nash equilibrium. Each player in the game wants to maximize their own payoff. This means that unilaterally changing the strategy of either player will not improve their payoff. At this point, a stable state is reached, and the combination of strategies between the two players constitutes the Nash equilibrium. For mixed strategies, it is impossible for a player to achieve a better payoff than the Nash equilibrium by changing only their own mixed strategy without changing the strategy of the other player. Therefore, the Nash equilibrium can be expressed as follows:

[0135]

[0136]

[0137] at this time and is the expected return of EV and SV at Nash equilibrium. According to the return matrix in step (4), the returns of EV and SV can be expressed as:

[0138]

[0139]

[0140] Among them E EV (p1,q1,q2) is the expected return of EV, E SV (p1,q1,q2) is the expected return of SV.

[0141] According to the expected returns of EV and SV, the Nash equilibrium can be further expressed as:

[0142]

[0143]

[0144] After solving the Nash equilibrium, we get (p1 * ,q1 * ,q2 * ) probability of the strategy, p1 * is the best response of EV to SV’s strategy, that is, the probability of EV changing lanes at Nash equilibrium; (q1 * ,q2 *) is the best response of SV to EV, that is, the probability of SV choosing various actions in Nash equilibrium. Therefore, when EV takes the action with probability p1 * Lane change, 1-p1 * Without changing lanes, SV takes q1 * Acceleration, q2 * Slowdown, 1-q1 * -q2 * At a constant speed, that is, at this time (p * ,q1 * ,q2 * ) corresponds to the best strategy combination of the lane-changing vehicle and the vehicle behind the target lane. Specifically, if p * >0.5, the EV changes lanes and the SV strategy is set to q1 * ,q2 * and q3 * The strategy corresponding to the maximum probability.

Claims

1. A lane-changing decision-making method based on game theory considering different vehicle types, characterized in that: The specific steps are as follows: Step (1) collecting vehicle information and surrounding vehicle information through the vehicle-mounted data collection device of the smart car; The road scene considered is a one-way two-lane road, which is straight and contains different types of vehicles. , the information set is described as: ； 、 and Represents vehicles The longitudinal position, longitudinal velocity and longitudinal acceleration at time t define the longitudinal position of the starting end of the current road =0, where longitudinal refers to the direction of vehicle travel, Indicates vehicle The lane where you are Indicates vehicle The type of vehicle is defined according to the vehicle length, which is Type I and Type II. Among them, vehicles with a length of less than 4.3 meters are Type I, and vehicles with a length of more than 5.9 meters are Type II; Step (2) making a lane change decision based on the collected information; vehicle According to the surrounding traffic conditions, including the speed and position information of surrounding vehicles, the vehicle is judged whether the current lane meets its own driving requirements. If not, the vehicle The initial lane-changing motivation will be generated, and the driver will further judge whether the road conditions in the adjacent lanes support the lane-changing behavior. If so, the driver will determine whether the lane-changing behavior is supported. The final lane-changing intention; otherwise, You can only continue driving in the original lane and do not need to make subsequent lane change decisions; the specific judgment method is as follows: Vehicle I is more interested in speed gains during driving, that is, it hopes to travel at a faster speed. When the vehicle ahead affects the acceleration of Vehicle I, Vehicle I will be motivated to change lanes, which means that the following conditions are met: ,and, ; in for Time Vehicle location, for time The position of the vehicle directly ahead, Indicates the minimum distance within which driving is not affected. Beyond this distance, the vehicle ahead will affect the driving decision. for Time Vehicle Speed ​​in the direction of travel, for time The speed of the vehicle directly ahead in the direction of travel; Unlike Model I, Model II not only pursues speed gains but also seeks stability, avoiding frequent changes in its own conditions due to lane changes. Therefore, Model II is more tolerant of vehicles ahead. It will only be motivated to change lanes if the vehicle ahead of it makes it difficult to maintain its current speed. This means that the following conditions are met: ,and, ; After the lane-changing motivation is met, it is also necessary to determine whether the road conditions in the adjacent lanes meet the lane-changing requirements; the adjacent lane of the current lane is called the target lane, and the first vehicle behind the target lane is called the lagging vehicle; the road conditions refer to whether the distance between the current vehicle and the lagging vehicle meets the safety conditions, so as to avoid a rear-end collision with the lagging vehicle in the target lane during the lane-changing process of the current vehicle; the safety distance must be guaranteed in the worst case, that is, the vehicle Even if emergency braking is applied, the lagging vehicle will not rear-end the vehicle; Vehicle Type I must meet the following requirements: ； in for The position of the vehicle's front end is always lagging behind. for The position of the front of the car, For vehicles The length of the body; , represents the distance between the two cars at time t, and further we can get: （1）； in and Respectively Time Vehicle and vehicles The emergency braking distance of the lagging vehicle is given by Sure, The vehicle's emergency braking acceleration; therefore, the vehicle The safe driving condition is that the distance between the two vehicles is not less than the difference between the braking distance of the rear vehicle and the braking distance of the vehicle itself. Considering that the vehicle type II is heavier than the vehicle type I and has poor emergency braking capability, a larger space should be required as a safe distance. The right side of formula (1) is called , and the final safety constraints are obtained: ； in It is called safety factor. The larger the safety factor, the higher the safety requirements of the vehicle. To be larger than Model I; After the lane change motivation and safety conditions are met, the vehicle conducts V2V communication to inform surrounding vehicles of the lane change request; Step (3) determine the game participants and establish a strategy combination for both parties; Using game theory methods in lane-changing decisions; vehicles It is necessary to play a game with the lagging vehicle in the target lane to improve the lane change success rate and reduce the occurrence of congestion. Therefore, the game participants are determined to be the lane-changing vehicle EV and the lagging vehicle SV in the target lane. The game between the two parties is represented by a four-tuple ; Where P is the set of game participants, P = {lane-changing vehicle (EV), lagging vehicle (SV)}, where the types of EV and SV are vehicle type I and vehicle type II; Refers to the set of optional actions of the game participants; the driving action set of EV is ={LC, DC}, LC means lane change, DC means no lane change; SV can choose to accept or reject the lane change behavior of EV. If it chooses to accept, it will be realized by decelerating or driving at a constant speed; if it chooses to reject, it will accelerate. Therefore, the driving action set of SV is ={UN, AC, DE}, UN means uniform speed, AC means acceleration, DE means deceleration; It is a set of strategy combinations of both parties in the game. There are six strategy combinations between the two parties in the game, so No lane change, acceleration); (No lane change, deceleration); (No lane change, constant speed); (Lane change, acceleration); (Lane change, deceleration); (Lane change, constant speed)}; is the set of payoffs of the game participants; Step (4) Analyze the benefits of both parties in the game; EV and SV achieve different returns when adopting different pure strategies and , where i={1,2}, j={1,2,3}; t={C,T} indicates that the vehicle types are model I and model II; the specific expression of the profit matrix is: ； Furthermore, different vehicle types have different payoffs. In the lane-changing game between EVs and SVs, both parties know the possible action sets of the other party. EVs and SVs will choose different actions with certain probability distributions based on different situations. Furthermore, the two players have different driving goals and pursuits. They both seek to maximize their own payoffs and choose the strategy that is most beneficial to them. Therefore, the game is a complete information non-cooperative mixed strategy game. In mixed strategy games, EV and SV randomly select different strategies with a certain probability. denote the probability of EV adopting two pure strategies: lane changing and no lane changing, respectively, and ; They represent the probability of SV adopting three pure strategies: acceleration, deceleration, and uniform speed, and According to the benefits of the two players adopting different strategies in the game and the probabilities of EV lane change and SV acceleration and deceleration, the payoff matrix of the mixed strategy case is expressed as: ； The expected return of the hybrid strategy of vehicle EV and vehicle SV is 、 , which is expressed as follows: ； ； Step (5) designing a profit function based on the factors that affect the vehicle's lane-changing behavior; The factors that affect the vehicle's lane-changing decision include the speed difference between the vehicle and the front and rear vehicles, and the distance between the vehicle and the front and rear vehicles. Therefore, when formulating the benefit function, the speed benefit is taken into account. and space benefits Set as the main decision factor of the target vehicle; In addition, vehicle type I pays attention to the type of vehicle in front and is unwilling to drive behind vehicle type II, so the benefit of the vehicle type in front is considered. ; Model II focuses on the stability of its own driving state, including speed stability and dimensional stability Different types of vehicles have different driving characteristics, different impacts on the road, and different corresponding benefit functions; the benefit of the current vehicle in step (4) is and the payoff of the lagging vehicle is The cases when t={C,T} are discussed separately; the benefits obtained when the current vehicle EV is model I and model II respectively and the vehicle behind the target lane SV is model I and model II respectively are analyzed; First, consider the case where the EV is model I and chooses to change lanes, mainly considering the speed gain. and space benefits ; Since vehicle type I is affected by the type of vehicle in front, the benefit of the vehicle type in front also needs to be considered If the lagging vehicle decides to accelerate or decelerate, the safety condition determined in step (2) changes, so security benefits need to be considered The benefits generated by the above conditions are specifically expressed as follows: ； ； ； in , Qualitatively indicates the type of vehicle ahead after lane change, Indicates the type of vehicle before changing lanes; is positive, Model II is negative, when EV changes lanes The benefit of lane changing is evaluated by comparing the vehicle types before and after the lane change; the speed benefit is specifically expressed as ,in Indicates that the vehicle ahead after changing lanes is The speed of time, Indicates this vehicle The speed of time, Indicates the road speed limit; the vehicle expects to achieve a higher driving speed. If the speed of the preceding vehicle is lower than its own speed, the benefit is negative, otherwise it is positive. The specific spatial benefit is expressed as ;in express The front position of the vehicle in front at that moment, express The front position of the vehicle at that moment, is the length of the vehicle in front; the greater the distance between the vehicle and the vehicle in front, the more operating space and reaction time the driver has, and the corresponding driving safety and comfort will be higher; the safety benefit is ;in is the speed of the vehicle behind after changing lanes; each benefit All are normalized and their values ​​satisfy , , , , , is the parameter that needs to be adjusted; Consider the case where the EV is a vehicle type II and chooses to change lanes. Except for not considering the benefit of the vehicle type in front, the other conditions are the same as those for the vehicle type I lane change, specifically expressed as: ； ； ； , , , , , , , , , is the parameter that needs to be adjusted; When the EV is Model I and chooses not to change lanes, the speed gain is still considered. and space benefits ; At this time, the type of vehicle in front of the EV does not change, so Since the EV does not change lanes, the vehicle behind Model I remains unchanged. Therefore, the benefit of Model I does not consider the impact of the vehicle behind it. When the EV is a Type II vehicle and chooses not to change lanes, in addition to considering the speed gain and space benefits In addition, spatial stability benefits must also be considered , where for The longitudinal position of the rear vehicle's front end at that moment, is the vehicle body length, Indicates safe distance; the larger the space behind the vehicle, the lower the chance of a rear-end collision. The greater this benefit, the better the space stability. Consider the case where the SV is vehicle type I and it chooses to accelerate. Vehicle type I is affected by the type of vehicle in front. If the EV changes lanes, the type of vehicle in front may change. ; When EV does not change lanes, it also needs to consider the benefits of the vehicle type in front , the benefits generated by the above conditions are specifically expressed as: ； ； Furthermore, the benefits when the SV is vehicle type I and chooses to decelerate are the same as in the above case; When SV is vehicle type II and acceleration is considered, vehicle type II expects its driving state to be stable, and the benefit is affected by speed changes. Speed ​​stability is defined as , where Indicates the speed before SV deceleration, Indicates the vehicle speed limit; excessive speed changes will affect the smooth driving of Model II; the benefits generated by the above conditions are specifically expressed as: ； ； In addition, the benefit when SV is vehicle type I and chooses to decelerate is the same as the above case; when SV is at a constant speed, the speed does not change, so it is not considered. ; Step (6) solve the Nash equilibrium and obtain the final driving decision; For mixed strategies, it is impossible for a participant to obtain a better payoff than the Nash equilibrium by changing only his own mixed strategy without changing the other party's strategy; therefore, the Nash equilibrium is expressed as follows: ； ； at this time and is the expected return of EV and SV at Nash equilibrium; according to the return matrix in step (4), the returns of EV and SV are expressed as: ； in is the expected return of EV, is the expected return of SV; According to the expected returns of EV and SV, the Nash equilibrium is further expressed as: ； ； After solving the Nash equilibrium, we get A strategy based on probability, is the best response of EV to the strategy adopted by SV, that is, the probability of EV changing lanes in Nash equilibrium; is the best response of SV to EV's strategy, that is, the probability of SV choosing various actions in Nash equilibrium; therefore, when EV takes the form of Change lanes, Without changing lanes, SV accelerate, slow down, When the speed is constant, it is the best strategy combination for the lane-changing vehicle and the vehicle behind in the target lane.

Images