A vehicle collaborative decision-making system and method based on incomplete information dynamic game

Abstract

The present invention relates to a vehicle collaborative decision-making system and method based on incomplete information dynamic game. The system includes an observation module, a driver intention recognition module, a belief pool updating module, a joint strategy benefit calculation module, and a behavior decision module; the observation module is connected to the driver intention recognition module and the joint strategy benefit calculation module, the driving intention recognition module is connected to the belief pool updating module, the belief pool updating module is connected to the joint strategy benefit calculation module, and the joint strategy benefit calculation module is connected to the behavior decision module; the method includes the following steps: step S1: judging the driving status information of both parties in a road right conflict through the observation module; step S2: judging the driving intention of the vehicle or driver based on the driving status information of the conflicting parties; step S3: updating the driver type distribution probability in the belief pool updating module; step S4: calculating the benefit function of all feasible joint strategies; step S5: deciding the most effective joint distribution strategy.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent traffic control, and specifically relates to a vehicle collaborative decision-making system and method based on incomplete information dynamic game. Background Art

[0002] Behavioral decision-making for smart cars is one of the key technologies for autonomous driving. Smart cars need to determine the optimal driving intention based on environmental information and driving status. During this process, the vehicle will interact with surrounding vehicles and even cause conflicts over right of way. Regarding the technical shortcomings of the above-mentioned existing technologies:

[0003] The invention patent with publication number CN112116100A discloses a technical solution for a game-theoretic decision-making method that takes driver types into consideration. This technical solution can improve the accuracy of the prediction of surrounding vehicle movements and the safety and feasibility of the intelligent vehicle's decision-making results. However, this method is not a collaborative decision-making method for multiple intelligent vehicles, and does not update the vehicle type distribution probability by observing the driving intentions of conflicting vehicles.

[0004] Patent publication number CN110962853A discloses a technical solution for a vehicle-to-vehicle (IoV) collaborative lane-changing method. This method improves lane-changing efficiency while ensuring traffic safety, thereby increasing road capacity. However, this method only works when both parties in the game can fully communicate and cannot resolve scenarios where the conflicting parties cannot communicate.

[0005] Patent publication number CN111267846A discloses a game-theory-based method for predicting the interactive behavior of surrounding vehicles. This method fully accounts for the uncertainty of movement trajectories caused by these interactions, making vehicle decision-making and planning safer and more reliable. However, this method fails to consider the calculation of benefits when complete information communication is incomplete, and cannot address the probability of different vehicle types adopting different strategies.

[0006] The invention patent with publication number CN111645692A discloses a technical solution for a method and system for identifying a driver's overtaking intention based on a hybrid strategy. Based on the overtaking habits of human drivers, it effectively solves the safety hazards during vehicle overtaking and improves the driving efficiency of the vehicle. However, this method only identifies the driving intention and does not provide a collaborative solution for meeting the equilibrium conditions.

[0007] In view of this, the present invention provides a vehicle collaborative decision-making system and method based on incomplete information dynamic game to solve the technical defects existing in the prior art. It uses incomplete information dynamic game to effectively describe the road right conflicts and behavioral interactions between vehicles, and obtains the collaborative decision-making solution with the highest expected benefit through a multi-objective benefit function. Summary of the Invention

[0008] The purpose of the present invention is to address the defects of the above-mentioned prior art and provide a vehicle collaborative decision-making system and method based on incomplete information dynamic game to solve the above-mentioned technical problems.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A vehicle collaborative decision-making system based on incomplete information dynamic game, comprising:

[0011] Observation module, driver intention recognition module, belief pool update module, joint strategy benefit calculation module, behavior decision module;

[0012] The observation module is connected to the driver intention recognition module and the joint strategy benefit calculation module, the driving intention recognition module is connected to the belief pool update module, the belief pool update module is connected to the joint strategy benefit calculation module, and the joint strategy benefit calculation module is connected to the behavior decision module;

[0013] The observation module is used to determine the driving status information of the two parties in the right-of-way conflict, and transmit the obtained driving status information of the two parties in the right-of-way conflict to the driver intention recognition module;

[0014] The driver intention recognition module is used to determine the driving intention data of the vehicle or driver based on the acquired driving status information, and transmit the determined driving intention data to the belief pool update module;

[0015] The belief pool update module is used to update the driver type distribution probability in the belief pool according to the driving intention data, calculate the new joint strategy distribution probability, and transmit the new joint strategy distribution probability to the joint strategy benefit calculation module;

[0016] The joint strategy benefit calculation module is used to calculate the benefit function of all feasible joint strategies, and calculate the expected benefit of the joint strategy in combination with the joint strategy distribution probability to obtain the optimal joint distribution strategy;

[0017] The behavior decision module is used to make behavior decisions for vehicles with right-of-way conflicts based on an optimal joint distribution strategy.

[0018] Preferably, the observation module is used to determine the driving status information of both parties in the right-of-way conflict, and the driving status information includes the vehicle's position information, speed information and acceleration information.

[0019] Preferably, the driver intention recognition module is used to determine the driving intention data of the vehicle or driver based on the acquired driving status information, and the driving intention data includes lane change intention data, acceleration intention data, deceleration intention data and parking intention data.

[0020] Preferably, the belief pool updating module includes driver type information and behavioral strategy distribution probabilities of different types of drivers, and the driver type information includes a tough type and a weak type.

[0021] Preferably, the probability update of the belief pool update module is adapted to the Bayesian equation. During the game process, when it is observed that the vehicle has taken a lane-merging or lane-maintaining action, the following posterior probability correction is performed:

[0022] Under the condition of merging behavior, the posterior probability of the car being a tough type is:

[0023]

[0024] Under the condition of merging behavior, the posterior probability of the vehicle being a weak type is:

[0025]

[0026] Under the condition of holding behavior, the posterior probability of the car being tough is:

[0027]

[0028] Under the condition of holding behavior, the posterior probability of the car being weak is:

[0029]

[0030] The sum of the updated vehicle type distribution probabilities is still 1:

[0031]

[0032] Among them, O1 is the distribution probability of the conflicting car being a tough type, O2 is the distribution probability of the conflicting car being a weak type, O1+O2=1;

[0033] π1 is the probability of merging if the conflicting vehicle is in a hard-line situation, and π2 is the probability of maintaining the lane if the conflicting vehicle is in a hard-line situation, π1+π2=1;

[0034] Ψ1 is the probability of merging when the conflicting vehicle is of the weak type, and Ψ2 is the probability of maintaining the lane when the conflicting vehicle is of the weak type, Ψ1+Ψ2=1.

[0035] As a preference, in the joint strategy benefit calculation module, the benefits of all feasible joint strategies are calculated based on the observation module and the belief pool update module:

[0036] Assume that in the belief pool of conflicting car 1, the distribution probability that conflicting car 2 is tough is Z1, and the distribution probability that conflicting car 2 is weak is Z2. The probability of allowing lane merging when conflicting car 2 is tough is The probability of refusing to merge is The probability of allowing the conflicting car to merge in the weak case is The probability of refusing to merge is

[0037] Under different joint strategy distributions, the benefits of conflict car 1 and conflict car 2 are expressed as express,

[0038] According to the profit matrix (see the table below) and distribution function,

[0039] Profit Matrix

[0040]

[0041] Calculate the expected returns of conflicting cars 1 and 2 when they choose different action strategies. The returns of conflicting cars 1 and 2 under different strategies are:

[0042]

[0043] Where Ω and Φ respectively describe the matrices of the probability distribution of each behavior, which are

[0044]

[0045]

[0046] in, They are the expected benefits of the conflicting vehicle 2 allowing and refusing to change lanes, and the expected benefits of the conflicting vehicle 1 maintaining and changing lanes.

[0047] Preferably, the behavior decision module obtains the expected benefits obtained by the conflicting car 2 and the conflicting car 1 adopting different strategies, and the decision set obtained by the Bayesian Nash equilibrium solution is calculated by the following formula:

[0048]

[0049]

[0050] in σ v ∈{0,1,2},σ p ∈{0,1,2}, respectively represent the acceleration of conflicting car 1 and conflicting car 2 and the behavior strategies adopted by each, and {0,,1,2} respectively represent {preparation, strategy 1, strategy 2}.

[0051] The present invention also provides a vehicle collaborative decision-making method based on incomplete information dynamic game, comprising the following steps:

[0052] Step S1: Determine the driving status information of both parties involved in the right-of-way conflict through the observation module, including the position data, speed data, and acceleration data of the vehicles;

[0053] Step S2: Determine the driving intention of the vehicle or driver based on the driving status information of the conflicting parties (conflicting vehicle 1 and conflicting vehicle 2), including lane change intention data, acceleration intention data, deceleration intention data, and parking intention data;

[0054] Step S3: Based on the driving intention, the driver type distribution probability in the belief pool update module is updated, and the new joint strategy distribution probability is calculated;

[0055] Step S4: Calculate the profit function of all feasible joint strategies, and calculate the expected profit of the joint strategy based on the distribution probability of the joint strategy;

[0056] Step S5: Determine the best joint distribution strategy to maximize the expected benefits of both conflicting parties.

[0057] The beneficial effect of this invention lies in that, when real-time, clear communication between vehicles is impossible and each vehicle is controlled by maximizing its own benefits, the right-of-way conflict game can be abstracted as an incomplete information, non-cooperative game. This incomplete information dynamic game effectively describes right-of-way conflicts and behavioral interactions between vehicles, and through a multi-objective benefit function, a collaborative decision-making solution with the highest expected benefit is obtained, thus avoiding conflicts between vehicles regarding right-of-way.

[0058] In addition, the present invention has a reliable design principle, a simple structure and a very broad application prospect.

[0059] It can be seen that compared with the prior art, the present invention has outstanding substantial features and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a principle block diagram of a vehicle collaborative decision-making system based on incomplete information dynamic game provided by the present invention.

[0061] Figure 2 It is the game tree of conflicting rooks under incomplete information.

[0062] Figure 3 It is the game tree of conflicting rook 2 under incomplete information.

[0063] Figure 4 This is a schematic diagram of the vehicle position changes when resolving road right conflicts according to this application.

[0064] Figure 5 This is a flow chart of a vehicle collaborative decision-making method based on incomplete information dynamic game provided by the present invention.

[0065] Among them, 1-Observation module, 2-Driver intention identification module, 3-Belief pool update module, 4-Joint strategy benefit calculation module, 5-Behavior decision module. DETAILED DESCRIPTION

[0066] The present invention will be described in detail below with reference to the accompanying drawings and through specific embodiments. The following embodiments are intended to explain the present invention, but the present invention is not limited to the following implementation modes.

[0067] Example 1:

[0068] like Figure 1-4 As shown, this embodiment provides a vehicle collaborative decision-making system based on incomplete information dynamic game, including:

[0069] Observation module 1, driver intention identification module 2, belief pool update module 3, joint strategy benefit calculation module 4, behavior decision module 5;

[0070] The observation module 1 is connected to the driver intention recognition module 2 and the joint strategy benefit calculation module 4, the driving intention recognition module 2 is connected to the belief pool update module 3, the belief pool update module 3 is connected to the joint strategy benefit calculation module 4, and the joint strategy benefit calculation module 4 is connected to the behavior decision module 5;

[0071] The observation module 1 is used to determine the driving status information of the two parties involved in the right-of-way conflict, and transmit the obtained driving status information of the two parties involved in the right-of-way conflict to the driver intention recognition module; the observation module 1 is used to determine the driving status information of the two parties involved in the right-of-way conflict, and the driving status information includes the vehicle's position information, speed information and acceleration information.

[0072] The driver intention recognition module 2 is used to judge the driving intention data of the vehicle or driver based on the acquired driving status information, and transmit the judged driving intention data to the belief pool update module; the driver intention recognition module 2 is used to judge the driving intention data of the vehicle or driver based on the acquired driving status information, and the driving intention data includes lane change intention data, acceleration intention data, deceleration intention data and parking intention data.

[0073] The belief pool update module 3 is used to update the driver type distribution probability in the belief pool according to the driving intention data, calculate the new joint strategy distribution probability, and transmit the new joint strategy distribution probability to the joint strategy benefit calculation module; the belief pool update module 3 includes the driver type information and the behavior strategy distribution probability of different types of drivers, and the driver type information includes a tough type and a weak type.

[0074] The joint strategy benefit calculation module 4 is used to calculate the benefit function of all feasible joint strategies, and calculate the expected benefit of the joint strategy in combination with the joint strategy distribution probability to obtain the optimal joint distribution strategy;

[0075] The behavior decision module 5 is used to make behavior decisions for vehicles with right-of-way conflicts based on an optimal joint distribution strategy.

[0076] The probability update of the belief pool update module 3 is adapted to the Bayesian equation. During the game process, when a vehicle is observed to have taken a lane-merging or lane-maintaining action, the following posterior probability correction is made:

[0077] Under the condition of merging behavior, the posterior probability of the car being a tough type is:

[0078]

[0079] Under the condition of merging behavior, the posterior probability of the vehicle being a weak type is:

[0080]

[0081] Under the condition of holding behavior, the posterior probability of the car being tough is:

[0082]

[0083] Under the condition of holding behavior, the posterior probability of the car being weak is:

[0084]

[0085] The sum of the updated vehicle type distribution probabilities is still 1:

[0086]

[0087] Among them, O1 is the distribution probability of the conflicting car being a tough type, O2 is the distribution probability of the conflicting car being a weak type, O1+O2=1;

[0088] π1 is the probability of merging if the conflicting vehicle is in a hard-line situation, and π2 is the probability of maintaining the lane if the conflicting vehicle is in a hard-line situation, π1+π2=1;

[0089] Ψ1 is the probability of merging when the conflicting vehicle is of the weak type, and Ψ2 is the probability of maintaining the lane when the conflicting vehicle is of the weak type, Ψ1+Ψ2=1.

[0090] In the joint strategy benefit calculation module 4, the benefits of all feasible joint strategies are calculated based on the observation module and the belief pool update module:

[0091] Assume that in the belief pool of conflicting car 1, the distribution probability that conflicting car 2 is tough is Z1, and the distribution probability that conflicting car 2 is weak is Z2. The probability of allowing lane merging when conflicting car 2 is tough is The probability of refusing to merge is The probability of allowing the conflicting car to merge in the weak case is The probability of refusing to merge is

[0092] Under different joint strategy distributions, the benefits of conflict car 1 and conflict car 2 are expressed as express,

[0093] According to the profit matrix (see the table below) and distribution function,

[0094] Profit Matrix

[0095]

[0096] Calculate the expected returns of conflicting cars 1 and 2 when they choose different action strategies. The returns of conflicting cars 1 and 2 under different strategies are:

[0097]

[0098] Where Ω and Φ respectively describe the matrices of the probability distribution of each behavior, which are

[0099]

[0100]

[0101] in, They are the expected benefits of the conflicting vehicle 2 allowing and refusing to change lanes, and the expected benefits of the conflicting vehicle 1 maintaining and changing lanes.

[0102] The behavior decision module obtains the expected benefits of conflicting cars 2 and 1 adopting different strategies. The decision set obtained through the Bayesian Nash equilibrium solution is calculated by the following formula:

[0103]

[0104]

[0105] in σ v ∈{0,1,2},σ p ∈{0,1,2}, respectively represent the acceleration of conflicting car 1 and conflicting car 2 and the behavior strategies adopted by each, and {0,1,2} respectively represent {preparation, strategy 1, strategy 2}.

[0106] In this embodiment, it is assumed that the profit function and the initial belief pool at a certain stage are known. Figure 2 、 Figure 3 , both parties in the game know their own type information, the probability distribution of the other party's type, the probability of choosing each strategy under different types, etc. The conflicting car 2 knows that its own type is weak and the payoffs under different strategies.

[0107] The expected returns of the conflicting vehicle 2 adopting the strategy of allowing and refusing to merge are respectively expressed as follows:

[0108]

[0109]

[0110] Obviously, the expected benefit of allowing lane change is higher than the benefit of refusing lane change, so the conflicting vehicle 2 will adopt the strategy of allowing lane change under the current working conditions. Similarly, the lane-changing vehicle knows the type and benefit of the conflicting vehicle 2 in addition to its own benefit function. Its game strategy tree is as follows: Figure 2 shown.

[0111] According to the game strategy tree of the merging car, the payoff of continuing to merge is:

[0112]

[0113] The payoff of adopting the stop-and-merge strategy is:

[0114]

[0115] Clearly, in this game, the merging smart car will choose to continue merging, while the conflicting car 2 will choose to allow it to change lanes. Using Bayes' theorem to calculate the probability distribution of merging vehicles within conflicting car 2, the posterior probability of conflicting car 1 being {0.6, 0.4} is obtained from the following equation:

[0116]

[0117]

[0118] Similarly, the updated type of conflicting car 2 is as follows: conflicting car 2 can obtain the posterior probability of conflicting car 1 as {0.4, 0.6}.

[0119]

[0120]

[0121] Repeating this process and summarizing the payoffs and belief pools for each cycle in the table below reveals that in this game, the conflicting vehicles will choose the Continue Merge strategy. With further calculations, the belief pool of conflicting vehicle 1 will converge to a weak strategy for conflicting vehicle 2, leading to a persistent Lane Change strategy for conflicting vehicle 1. In the belief pool of conflicting vehicle 2, the strategy for conflicting vehicle 1 will gradually converge to a strong strategy, allowing it to change lanes. Under this scenario, neither vehicle has an incentive to change its strategy, so {Continue Merge, Allow Merge} forms a subgame-perfect Bayesian Nash equilibrium.

[0122] The benefits, behaviors and belief pool information of the vehicles on both sides of the conflict are updated.

[0123]

[0124] In fact, in each stage of the game in this example, two Nash equilibria exist: {Merge, Allow Merge} and {Return, Refuse Merge}. That is, in the current payoff function, when conflicting car 1 chooses to merge, conflicting car 2 chooses to allow the merge; when conflicting car 2 chooses to return to its original lane, conflicting car 1 chooses to refuse the merge. During the game, neither player has any incentive to change their beliefs or action strategies, thus satisfying a refined Bayesian Nash equilibrium.

[0125] Figure 4 This is a diagram of the position changes under the behavioral control based on the current game decision. The numbers in the figure indicate the positions of the two vehicles at different sampling times. Conflicting vehicle 1 will continue to choose to change lanes, while conflicting vehicle 2 will continue to choose to allow lane changes, thereby slowing down and allowing conflicting vehicle 1 to enter in front of it.

[0126] Example 2:

[0127] like Figure 5As shown, this embodiment provides a vehicle collaborative decision-making method based on incomplete information dynamic game, including the following steps:

[0128] Step S1: Determine the driving status information of both parties involved in the right-of-way conflict through the observation module, including the position data, speed data, and acceleration data of the vehicles;

[0129] Step S2: Determine the driving intention of the vehicle or driver based on the driving status information of the conflicting parties (conflicting vehicle 1 and conflicting vehicle 2), including lane change intention data, acceleration intention data, deceleration intention data, and parking intention data;

[0130] Step S3: Based on the driving intention, the driver type distribution probability in the belief pool update module is updated, and the new joint strategy distribution probability is calculated;

[0131] Step S4: Calculate the profit function of all feasible joint strategies, and calculate the expected profit of the joint strategy based on the distribution probability of the joint strategy;

[0132] Step S5: Determine the best joint distribution strategy to maximize the expected benefits of both conflicting parties.

[0133] The above disclosure is only a preferred embodiment of the present invention, but the present invention is not limited thereto. Any non-creative changes that can be thought of by those skilled in the art, as well as several improvements and modifications made without departing from the principles of the present invention, should fall within the scope of protection of the present invention.

Claims

1. A vehicle collaborative decision-making system based on incomplete information dynamic game, characterized by: include: Observation module, driver intention recognition module, belief pool update module, joint strategy benefit calculation module, behavior decision module; The observation module is connected to the driver intention recognition module and the joint strategy benefit calculation module, the driver intention recognition module is connected to the belief pool update module, the belief pool update module is connected to the joint strategy benefit calculation module, and the joint strategy benefit calculation module is connected to the behavior decision module; The observation module is used to determine the driving status information of the two parties in the right-of-way conflict, and transmit the obtained driving status information of the two parties in the right-of-way conflict to the driver intention recognition module; The driver intention recognition module is used to determine the driving intention data of the vehicle or driver based on the acquired driving status information, and transmit the determined driving intention data to the belief pool update module; The belief pool update module is used to update the driver type distribution probability in the belief pool according to the driving intention data, calculate the new joint strategy distribution probability, and transmit the new joint strategy distribution probability to the joint strategy benefit calculation module; The joint strategy benefit calculation module is used to calculate the benefit function of all feasible joint strategies, and calculate the expected benefit of the joint strategy in combination with the joint strategy distribution probability to obtain the optimal joint distribution strategy; The behavior decision module is used to make behavior decisions for vehicles with right-of-way conflicts based on the optimal joint distribution strategy; In the joint strategy benefit calculation module, the benefits of all feasible joint strategies are calculated based on the observation module and the belief pool update module: Assume that in the belief pool of conflicting car 1, the distribution probability that conflicting car 2 is tough is Z1, and the distribution probability that conflicting car 2 is weak is Z2. The probability of allowing lane merging when conflicting car 2 is tough is , the probability of refusing to merge is , + =1; in the case of the weak type, the probability of allowing the lane change is Γ1, and the probability of refusing the lane change is Γ2, Γ1+Γ2=1; Under different joint strategy distributions, the benefits of conflict car 1 and conflict car 2 are expressed as express, According to the payoff matrix and distribution function, the expected payoffs of conflicting cars 1 and 2 when they choose different action strategies are calculated. The payoffs of conflicting cars 1 and 2 under different strategies are: in and The matrices that describe the probability distribution of each behavior are in, The expected benefits of the conflicting vehicle 2 taking the lane change policy and the lane change policy, and the expected benefits of the conflicting vehicle 1 taking the lane change policy and the lane change policy, respectively; The behavior decision module obtains the expected benefits of conflicting cars 2 and 1 adopting different strategies. The decision set obtained through the Bayesian Nash equilibrium solution is calculated by the following formula: in , respectively represent the acceleration of conflicting car 1 and conflicting car 2 and the behavioral strategies adopted by each, and {0, 1, 2} respectively represent {prepare, strategy 1, strategy 2}.

2. The vehicle collaborative decision-making system based on incomplete information dynamic game according to claim 1 is characterized in that: The observation module is used to determine the driving status information of both parties in the right-of-way conflict, and the driving status information includes the vehicle's position information, speed information and acceleration information.

3. The vehicle collaborative decision-making system based on incomplete information dynamic game according to claim 2 is characterized in that: The driver intention recognition module is used to determine the driving intention data of the vehicle or driver based on the acquired driving status information. The driving intention data includes lane change intention data, acceleration intention data, deceleration intention data and parking intention data.

4. The vehicle collaborative decision-making system based on incomplete information dynamic game according to claim 3 is characterized in that: The belief pool updating module includes driver type information and behavior strategy distribution probabilities of different types of drivers. The driver type information includes a tough type and a weak type.

5. The vehicle collaborative decision-making system based on incomplete information dynamic game according to claim 4 is characterized in that: The probability update of the belief pool update module is adapted to the Bayesian equation. During the game, when a vehicle is observed to have taken a lane-merging or lane-maintaining action, the following posterior probability correction is made: Under the condition of merging behavior, the posterior probability of the car being a tough type is: ； Under the condition of merging behavior, the posterior probability of the vehicle being a weak type is: ； Under the condition of holding behavior, the posterior probability of the car being tough is: ； Under the condition of holding behavior, the posterior probability of the car being weak is: ； The sum of the updated vehicle type distribution probabilities is still 1: ； Among them, O1 is the distribution probability of the conflicting car being a tough type, O2 is the distribution probability of the conflicting car being a weak type, O1+O2=1; π 1 is the probability of the conflicting vehicle changing lanes in a tough situation, π 2 is the probability of the conflict car maintaining the tough type, π 1+ π 2=1; Ψ 1 is the probability of the conflicting vehicle merging into a weak lane. Ψ 2 is the probability of the conflict car being in a weak state. Ψ 1+ Ψ 2=1.

6. A vehicle collaborative decision-making method based on incomplete information dynamic game, characterized in that: The following steps are involved: Step S1: Determine the driving status information of both parties involved in the right-of-way conflict through the observation module, including the position data, speed data, and acceleration data of the vehicles; Step S2: judging the driving intention of the vehicle or driver based on the driving status information of the conflicting parties, including lane change intention data, acceleration intention data, deceleration intention data, and parking intention data; Step S3: Based on the driving intention, the driver type distribution probability in the belief pool update module is updated, and the new joint strategy distribution probability is calculated; Step S4: Calculate the profit function of all feasible joint strategies, and calculate the expected profit of the joint strategy based on the distribution probability of the joint strategy; Step S5: Determine the best joint distribution strategy to maximize the expected benefits of both conflicting parties; In the joint strategy benefit calculation module, the benefits of all feasible joint strategies are calculated based on the observation module and the belief pool update module: Assume that in the belief pool of conflicting car 1, the distribution probability that conflicting car 2 is tough is Z1, and the distribution probability that conflicting car 2 is weak is Z2. The probability of allowing lane merging when conflicting car 2 is tough is , the probability of refusing to merge is , + =1; in the case of the weak type, the probability of allowing the lane change is Γ1, and the probability of refusing the lane change is Γ2, Γ1+Γ2=1; Under different joint strategy distributions, the benefits of conflict car 1 and conflict car 2 are expressed as express, According to the payoff matrix and distribution function, the expected payoffs of conflicting cars 1 and 2 when they choose different action strategies are calculated. The payoffs of conflicting cars 1 and 2 under different strategies are: in and The matrices that describe the probability distribution of each behavior are in, The expected benefits of the conflicting vehicle 2 taking the lane change policy and the lane change policy, and the expected benefits of the conflicting vehicle 1 taking the lane change policy and the lane change policy, respectively; The behavior decision module obtains the expected benefits of conflicting cars 2 and 1 adopting different strategies. The decision set obtained through the Bayesian Nash equilibrium solution is calculated as follows: in , respectively represent the acceleration of conflicting car 1 and conflicting car 2 and the behavioral strategies adopted by each, and {0, 1, 2} respectively represent {prepare, strategy 1, strategy 2}.

Images