Intelligent automobile human-like decision-making method and system based on virtual-real game double-system parallel

Abstract

The application discloses an intelligent automobile human-like decision-making method and system based on a virtual-real game double-system parallel, and the method comprises the following steps: an intelligent automobile establishes a virtual game system with surrounding vehicles according to self-cognitive knowledge and understanding of the types of the surrounding vehicles; when the surrounding vehicles make actual observations and optimal responses according to the strategy adopted by the intelligent automobile, the surrounding vehicles establish a real game system with the intelligent automobile; the intelligent automobile acquires the actual optimal response strategy of the surrounding vehicles, verifies the deviation between the optimal response strategy adopted by the surrounding vehicles in the virtual game system and the cognitive optimal response strategy, and corrects the understanding of the types of the surrounding vehicles and refreshes the belief according to the deviation, so as to form new cognition of the types of the surrounding vehicles; in the next stage of double-system parallel operation, the above steps are repeatedly executed, the intelligent automobile makes corresponding human-like decisions based on the new cognition, and the double-system cognition converges until the merging goal of the intelligent automobile is achieved; and the method improves the human-like level of the decision-making of the intelligent automobile.

Description

Technical Field

[0001] This invention relates to the technical field of human-like decision-making in intelligent vehicles, and particularly to a human-like decision-making method and system for intelligent vehicles based on a dual-system approach of virtual and real game theory. Background Technology

[0002] Intelligent vehicle technology is at the forefront and a hot topic in the future development of automotive technology. Intelligent vehicles are those equipped with advanced sensing devices, onboard computing units, and control execution modules, possessing capabilities such as perception, cognition, decision-making, planning, and control execution. They can achieve safe, energy-efficient, comfortable, and economical driving, and ultimately, can mimic human driving to achieve autonomous driving functions. The widespread adoption of intelligent vehicles is expected to improve road safety, optimize traffic efficiency, and enhance the travel experience for drivers and passengers.

[0003] The decision-making layer is the central hub of intelligent driving technology, playing a role similar to the central nervous system of a human driver's brain. It uses the rich environmental information obtained by sensing devices to issue instructions to the control execution module according to the driving task and control objectives, and is the key to affecting the effectiveness of intelligent vehicles in completing driving tasks and the process experience.

[0004] Experienced human drivers possess a deep understanding of complex traffic information, vehicle motion characteristics, and multi-vehicle interaction mechanisms, exhibiting excellent self-learning, self-evolution, and self-upgrading capabilities. This provides a natural reference for designing human-like decision-making methods for intelligent vehicles. Therefore, further enhancing the human-like nature of decision-making is a key focus in designing high-quality human-like decision-making methods for intelligent vehicles and an important indicator for evaluating the effectiveness of intelligent vehicle decision-making.

[0005] However, current research typically designs decision-making methods from a single dimension, treating the participants in the environment in which the intelligent vehicle operates as fixed obstacles. The intelligent vehicle's decision-making layer needs to make unilateral adaptive decisions based on these obstacles, which lacks consideration from a micro perspective and ignores the continuous interaction characteristics at a deeper level. The level of human-like decision-making needs to be improved. Summary of the Invention

[0006] In view of this, the present invention provides a human-like decision-making method for intelligent vehicles based on a dual-system parallel virtual-real game, which can solve the problems of single-dimensional design, neglect of interactive characteristics, and the need to improve the human-like level of decision-making in intelligent vehicles.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, embodiments of the present invention provide a human-like decision-making method for intelligent vehicles based on a dual-system parallel virtual-real game theory approach, comprising the following steps:

[0009] S1. Intelligent vehicles establish a virtual game system with surrounding vehicles based on their own cognitive knowledge and understanding of the types of surrounding vehicles.

[0010] S2. After the surrounding vehicles make actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, the surrounding vehicles establish a real-world game system with the intelligent vehicle.

[0011] S3. The intelligent vehicle obtains the actual optimal response strategies of surrounding vehicles and verifies the discrepancy between these strategies and the optimal response strategies perceived by the surrounding vehicles in the virtual game system.

[0012] S4. Intelligent vehicles correct their understanding of the types of surrounding vehicles based on deviations and refresh their beliefs, forming a new perception of the types of surrounding vehicles.

[0013] S5. In the next stage of parallel operation of the dual systems, repeat steps S1-S4 above. The intelligent vehicle makes corresponding human-like decisions based on the new cognition until the cognition of the dual systems converges and the intelligent vehicle achieves its lane-merging goal. The dual systems consist of a virtual game system and a real game system.

[0014] Further, step S1 includes:

[0015] S101. The intelligent vehicle estimates the type of surrounding vehicles based on the lateral and longitudinal driving behaviors and driving information of surrounding vehicles in the target lane during the recorded historical interaction time domain; the types include: aggressive, moderate and conservative.

[0016] S102. According to the sequential game rules, the intelligent vehicle itself is positioned as the forerunner in the sequential game, and the surrounding vehicles are positioned as the follower in the sequential game.

[0017] S103. Based on the intelligent vehicle's estimation of the types of surrounding vehicles and the sequential game action allocation, predict the response strategies of surrounding vehicles one by one for all the optional strategies in the intelligent vehicle's own game strategy set.

[0018] S104. Based on the game objective function of the intelligent vehicle and the game objective function of the surrounding vehicles, and based on the type of intelligent vehicle and the type of estimation of the surrounding vehicles, construct a virtual sequential game model.

[0019] S105. The problem of solving the virtual sequential game model is transformed into a model optimization problem. The virtual sequential game model is optimized and solved by reverse traversal and pruning of the sequential game expansion tree. The strategy pair under the game equilibrium state reached by the intelligent vehicle and the surrounding vehicles in the subjective perception of the intelligent vehicle is output so that the intelligent vehicle can make actual decisions.

[0020] Furthermore, the process of constructing the game objective function of the intelligent vehicle in step S104 is as follows:

[0021] Based on the rules of sequential game theory, and considering the arbitrary lateral acceleration strategies that intelligent vehicles can choose... And any longitudinal acceleration strategy selectable by surrounding vehicles Constructing the game objective function for intelligent vehicles ;

[0022] Among them, the game objective function of intelligent vehicles This includes: the objective function of lane merging efficiency game, the objective function of right-of-way attribution game, and the objective function of driving safety game;

[0023] 1) The objective function of the parallel track efficiency game is as follows:

[0024]

[0025] in, This represents the single-stage duration of a sequential game. This represents the Y-coordinate of the target lane centerline. This indicates the Y-coordinate of the current lane centerline;

[0026] 2) The objective function of the right-of-way attribution game is as follows:

[0027]

[0028] in, Indicating that smart cars adopt strategies It crosses the lane divider, gains road priority, and sets the payoff of the intelligent vehicle's right-of-way attribution game objective function to 1; Indicating that smart cars adopt strategies However, since the lane divider has not yet been crossed, the right of way still belongs to the surrounding vehicles in the target lane, so the payoff of the intelligent vehicle's right-of-way attribution game objective function is set to 0.

[0029] 3) The objective function for the driving safety game is as follows:

[0030]

[0031] in, This represents the initial coordinates of the intelligent vehicle. Indicates the initial coordinates of surrounding vehicles. This indicates the longitudinal speed of the intelligent vehicle. This indicates the longitudinal speed of surrounding vehicles.

[0032] Furthermore, the objective function construction process for the surrounding vehicle game in step S104 is as follows:

[0033] Based on the rules of sequential game theory, and considering the arbitrary lateral acceleration strategies that intelligent vehicles can choose... And any longitudinal acceleration strategy selectable by surrounding vehicles Construct the game objective function of the surrounding vehicles. ;

[0034] Among them, the objective function of the surrounding vehicles in the game. This includes: the objective function of driving safety game, the objective function of right-of-way attribution game, and the objective function of driving disturbance game;

[0035] a) The objective function for the driving safety of surrounding vehicles is as follows:

[0036] The driving safety objective for surrounding vehicles is determined by the strategies adopted by both surrounding vehicles and the intelligent vehicle from their respective policy sets. , The objective function of the game of driving safety of surrounding vehicles induced by this. Equivalent to the objective function of the driving safety game for intelligent vehicles ;

[0037] b) The objective function of the right-of-way attribution game for surrounding vehicles is as follows:

[0038]

[0039] in, Indicating that smart cars adopt strategies And cross the lane dividing line, surrounding vehicles lose their right-of-way, and the objective function payoff of the right-of-way attribution game for surrounding vehicles is set to -1; Indicating that smart cars adopt strategies However, since the lane divider has not been crossed, the surrounding vehicles have not improved their performance, so the payoff of the right-of-way attribution game objective function for the surrounding vehicles is set to 0.

[0040] c) The objective function for the game of interference from surrounding vehicles is as follows:

[0041] The driving interference of surrounding vehicles is measured by the relationship between the interference timeliness and the ideal timeliness of surrounding vehicles; interference timeliness is defined as the timeliness of surrounding vehicles in taking strategies. The time required to reach the current longitudinal position of the intelligent vehicle; ideal timeliness is defined as the time required for surrounding vehicles to reach the current longitudinal position of the intelligent vehicle while maintaining their original speed without engaging in game-like interactive decision-making with the intelligent vehicle; the driving interference target includes promoting interference and suppressing interference;

[0042]

[0043]

[0044]

[0045]

[0046] in, The objective function represents the game objective function for dealing with the interference of surrounding vehicles. This indicates that when surrounding vehicles adopt strategies within the acceleration range from their strategy set in order to cope with the lateral lane-changing behavior of the intelligent vehicle, it is considered to promote interference timeliness, which is less than ideal timeliness. This indicates that when surrounding vehicles adopt a strategy within the deceleration range from their strategy set in order to cope with the lateral lane-changing behavior of the intelligent vehicle, it belongs to the timeliness of interference suppression, which is greater than the ideal timeliness. It indicates ideal timeliness.

[0047] Furthermore, in step S104, the virtual sequential game model is constructed as follows:

[0048]

[0049]

[0050] in, This represents a virtual game model for intelligent vehicles; the type of intelligent vehicle is denoted as... , ∈{radical, moderate, conservative}, corresponding weight factors are ;

[0051] This represents a virtual game model of surrounding vehicles in the subjective perception of an intelligent vehicle; the types of surrounding vehicles in the subjective perception of an intelligent vehicle are: , The weighting factor is .

[0052] Further, in step S105, the problem of solving the virtual sequential game model is transformed into a model optimization problem, including:

[0053]

[0054]

[0055]

[0056] in, , These represent the strategies predicted by the intelligent vehicle for surrounding vehicles. , Response strategies; This represents the set of strategies for surrounding vehicles. For elements in the policy set; This represents the strategy pair in the game equilibrium state reached by the intelligent vehicle in its output and the surrounding vehicles in the intelligent vehicle's subjective perception.

[0057] Further, step S2 includes:

[0058] S201. After the surrounding vehicles have made actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, the surrounding vehicles then respond according to their own actual types. To implement decision-making strategies In order to make the best practical response;

[0059] S202. The intelligent vehicle establishes a realistic game model with surrounding vehicles as follows:

[0060]

[0061]

[0062] in, This represents a real-world game theory model for intelligent vehicles. ∈{radical, moderate, conservative}, corresponding weight factors are ; This represents the actual decisions made by the intelligent vehicle based on the virtual game system. Represents a real-world game model involving surrounding vehicles;

[0063] S203. Transform the problem of solving the real-world game model into a model optimization problem:

[0064]

[0065] Among them, intelligent vehicles, which occupy the position of first mover in the sequential game, have already made practical decisions. It only requires optimizing the actual decision-making strategies already made by smart cars. As parameters, any response strategy that a real-type surrounding vehicle might take from its policy set. This is an optimization problem with variables; the output should be the actual optimal response strategy for the surrounding vehicles of the true type. And based on this, the surrounding vehicles make the best actual response decisions.

[0066] Further, step S3 includes:

[0067] Intelligent vehicles learn the actual optimal response strategies of surrounding vehicles. This confirms the optimal response strategy adopted by surrounding vehicles as perceived in the virtual game system. The resulting deviation between the payoffs of the self-game objective function:

[0068]

[0069] in, This represents the deviation of the objective function payoff of intelligent vehicles in virtual and real-world game systems.

[0070] Furthermore, the payoff deviation caused by the intelligent vehicle's response strategy based on the verified optimal response strategies of surrounding vehicles and the optimal response strategy adopted in the virtual game system's perception. Correcting one's understanding of the types of vehicles around them, refreshing one's beliefs about intelligent vehicles, and forming a new understanding of the types of vehicles around them. .

[0071] Secondly, embodiments of the present invention also provide an intelligent vehicle human-like decision-making system based on a dual-system parallel virtual-real game theory approach, comprising:

[0072] The subjective decision-making module is used by intelligent vehicles to establish a virtual game system with surrounding vehicles based on their own cognitive knowledge and understanding of the types of surrounding vehicles.

[0073] The objective decision-making module is used to establish a realistic game system between the surrounding vehicles and the intelligent vehicle after the surrounding vehicles have made actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle.

[0074] The strategy verification module is used by intelligent vehicles to obtain the actual optimal response strategies of surrounding vehicles and verify the deviation between these strategies and the optimal response strategies perceived by the virtual game system.

[0075] The understanding refresh module is used by intelligent vehicles to correct their understanding of the types of surrounding vehicles based on deviations and refresh their beliefs to form a new cognition of the types of surrounding vehicles.

[0076] In the next stage of parallel operation of the dual systems, the above modules are repeatedly executed, and the intelligent vehicle makes corresponding human-like decisions based on the new cognition until the cognition of the dual systems converges and the intelligent vehicle achieves its lane-merging goal; the dual systems consist of a virtual game system and a real game system.

[0077] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a human-like decision-making method for intelligent vehicles based on a dual-system parallel virtual-real game. This method addresses the deviation between subjective and objective decision-making and promotes the convergence of deviations through the parallel integration of the dual systems. While deepening the intelligent vehicle's understanding of the types of surrounding vehicles, it improves the human-like level of intelligent vehicle decision-making through the gradual construction of anthropomorphic cognitive beliefs. Attached Figure Description

[0078] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0079] Figure 1 A flowchart of a human-like decision-making method for intelligent vehicles based on a dual-system approach of virtual and real game theory;

[0080] Figure 2 This is a schematic diagram illustrating the parallel operation of a virtual and real game-themed dual system during high-speed lane merging.

[0081] Figure 3 A schematic diagram of the overall principle of a human-like decision-making method for intelligent vehicles based on a dual-system approach of virtual and real game theory;

[0082] Figure 4 A block diagram of an intelligent vehicle human-like decision-making system based on a dual-system approach of virtual and real game theory;

[0083] Figure 5 A flowchart of a virtual game system;

[0084] Figure 6 This is a flowchart of a real-world game system. Detailed Implementation

[0085] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0086] The technical solution of this invention treats intelligent vehicles and environmental participants as flexible intelligent agents with autonomous capabilities, and deeply integrates the multi-agent game interaction mechanism into the design process of intelligent vehicle human-like decision-making methods. This is conducive to improving the human-like level of intelligent vehicles, and helps to achieve high-quality collaboration and harmonious interaction between human wisdom and machine intelligence. It has important positive significance for building an intelligent vehicle technology innovation system, and ultimately realizing the vision of a safe, green, efficient, and civilized intelligent vehicle powerhouse.

[0087] In high-speed lane-changing situations, intelligent vehicles cannot fully grasp information such as the types of surrounding vehicles. They need to make subjective decisions based on their own cognitive knowledge and the surrounding vehicles. In this subjective decision-making, intelligent vehicles need to make an estimate of the types of surrounding vehicles and then predict the coping actions that surrounding vehicles of that type may take in response to their own strategies. At the same time, surrounding vehicles will respond to the strategies already taken by the intelligent vehicle with their actual types, thus forming an objective decision with the intelligent vehicle.

[0088] Based on this, this invention discloses a human-like decision-making method for intelligent vehicles based on a dual-system parallel approach of virtual and real game theory, in order to address the deviation between subjective and objective decision-making. By integrating the dual systems in parallel, the method promotes the convergence of deviations, thereby deepening the intelligent vehicle's understanding of the types of surrounding vehicles and further improving the human-like level of intelligent vehicle decision-making.

[0089] Reference Figure 1 As shown, it specifically includes:

[0090] S1. The intelligent vehicle establishes a virtual game system with the surrounding vehicles based on its own cognitive knowledge and understanding of the types of surrounding vehicles; among them, the main vehicle with the intention to actively change lanes is the intelligent vehicle, and the vehicles that are freely moving in front of and behind the target lane are the surrounding vehicles.

[0091] S2. After the surrounding vehicles make actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, the surrounding vehicles establish a real-world game system with the intelligent vehicle.

[0092] S3. The intelligent vehicle obtains the actual optimal response strategies of surrounding vehicles and verifies the discrepancy between these strategies and the optimal response strategies perceived by the surrounding vehicles in the virtual game system.

[0093] S4. Intelligent vehicles correct their understanding of the types of surrounding vehicles based on deviations and refresh their beliefs, forming a new perception of the types of surrounding vehicles.

[0094] S5. In the next stage of parallel operation of the dual systems, repeat steps S1-S4 above. The intelligent vehicle makes corresponding human-like decisions based on the new cognition until the cognition of the dual systems converges and the intelligent vehicle achieves its lane-merging goal. The dual systems consist of a virtual game system and a real game system.

[0095] Both the virtual and real-world game systems established above are based on sequential game theory, such as... Figure 2As shown, the intelligent vehicle interacting on the virtual end and the surrounding vehicles together constitute a virtual game system, while the intelligent vehicle interacting on the ground end and the surrounding vehicles together constitute a real game system. The intelligent vehicle in the virtual game system and the real game system are the same entity. The surrounding vehicles in the real game system are the actual types of the surrounding vehicles. This type is knowledge possessed unilaterally by the surrounding vehicle entity and is unobservable to the intelligent vehicle. The surrounding vehicles in the virtual game system are the intelligent vehicle's understanding of the types of surrounding vehicles based on its own cognitive knowledge. This cognitive knowledge may have subjective limitations and uncertainties, and therefore may be consistent with the actual types of surrounding vehicles or may be different from the actual types of surrounding vehicles.

[0096] like Figure 2 As shown, decision scenario 1: dark blue Intelligent vehicles in virtual and real-world game systems; red in the real-world game system The actual surrounding vehicles are of an aggressive type; the light green color in the virtual game system... The intelligent vehicle perceives the surrounding vehicles as conservative in its subjective understanding; in decision-making scenario 1, the intelligent vehicle's knowledge of the surrounding vehicles has a significant bias.

[0097] Decision Scenario 2: Dark Blue Intelligent vehicles in virtual and real-world game systems; black players in real-world game systems. The actual surrounding vehicles are of moderate type; the light black area in the virtual game system... In the intelligent vehicle's subjective perception of surrounding vehicles, the intelligent vehicle perceives them as of the moderate type; in decision scenario 2, the intelligent vehicle's knowledge of surrounding vehicles has a small bias.

[0098] Decision Scenario 3: Dark Blue Intelligent vehicles in virtual and real-world gaming systems; deep green in real-world gaming systems. The actual surrounding vehicles are of a conservative type; the light red ones in the virtual game system... In the intelligent vehicle's subjective perception of surrounding vehicles, the intelligent vehicle perceives them as aggressive; in decision scenario 3, the intelligent vehicle's knowledge of surrounding vehicles has a significant bias.

[0099] Decision Scenario 4: Dark Blue Intelligent vehicles in virtual and real-world game systems; gray areas in real-world game systems. The actual surrounding vehicles can be of any type among aggressive, moderate, and conservative; the light gray area in the virtual game system... For the surrounding vehicles in the subjective perception of the intelligent vehicle, the intelligent vehicle recognizes them as any type among aggressive, moderate, and conservative. In decision scenario 4, the intelligent vehicle's knowledge of the surrounding vehicles may have a large bias or a small bias.

[0100] It is evident that this cognitive knowledge may be subject to subjective limitations and uncertainties, and thus may be consistent with the actual types of vehicles in the surrounding area, or may differ from them. Therefore, this cognition is a stochastic process, which expands the scope of applicability of the present invention.

[0101] The following is combined with Figure 3 The following is a detailed explanation of each of the above steps:

[0102] 1. Step S1 constructs a subjective decision-making virtual game system, specifically including steps S101~S105;

[0103] Step S101 aims to enable the intelligent vehicle to estimate the types of surrounding vehicles in the target lane for merging, specifically categorized as aggressive, moderate, or conservative. Accurately understanding the types of the surrounding intelligent agents with which the vehicle continuously interacts has a significant impact on the intelligent vehicle's ability to make rational and effective decisions. For example, when surrounding vehicles are aggressive, they may accelerate to contest the intelligent vehicle's lane-merging decision, potentially causing serious conflict if the intelligent vehicle blindly merges. When surrounding vehicles are conservative, they may slow down and compromise even with small lateral displacements, requiring the intelligent vehicle to merge into the target lane as quickly as possible to minimize disruption to highway traffic caused by hesitant decision-making.

[0104] Intelligent vehicles estimate and make preliminary judgments about the types of surrounding vehicles based on the lateral and longitudinal driving behaviors and driving information of surrounding vehicles recorded in the historical interaction time domain. This involves all acceleration, speed, heading angle and absolute coordinate information of surrounding vehicles in the historical time domain H.

[0105] Step S102 involves sequential game allocation, aiming to assign the action positions of the intelligent vehicle and surrounding vehicles in the sequential game. The virtual game system involved in step S1 is based on sequential game design. The intelligent vehicle's intention to change lanes is a prerequisite for potential conflict with surrounding vehicles in the target lane. The intelligent vehicle's actions are leading, and the actions of surrounding vehicles are responsive. This invention positions the intelligent vehicle as the leader in the sequential game and the surrounding vehicles as followers, realizing the allocation of actions between the intelligent vehicle and surrounding vehicles in the subjective decision-making virtual game system.

[0106] Step S10 involves action prediction, aiming to predict the possible response strategies of surrounding vehicles. Specifically, the sequential game process involves the intelligent vehicle taking the first action determining its optimal strategy in advance based on the prediction of the actions taken by surrounding vehicles following its own strategy. The surrounding vehicles following its own actions then further determine their optimal response strategies based on the strategies already adopted by the intelligent vehicle taking the first action. Before formally making a lane-merging decision, the intelligent vehicle, based on the aforementioned estimation and preliminary judgment of the types of surrounding vehicles and the sequential game action allocation, predicts the response strategies of surrounding vehicles one by one for all available strategies in its own game strategy set.

[0107] The virtual sequential game model in step S104 is the core of autonomous decision-making for intelligent vehicles. This invention defines the key elements of the sequential game as: game players. , For smart cars, For surrounding vehicles; strategic space , A set of strategies for intelligent vehicles, in which A subset or element of the strategy set for intelligent vehicles, representing any lateral movement that the intelligent vehicle can choose (e.g., ...). Figure 2 The acceleration strategy is shown in the Y direction (the direction perpendicular to the forward direction). The strategy set for surrounding vehicles, where A subset or element of the strategy set of surrounding vehicles, representing any possible longitudinal direction that surrounding vehicles can choose (e.g., ...). Figure 2 The acceleration strategy is shown in the X direction (forward direction).

[0108] set up The objective function of the game for intelligent vehicles depends on the strategies of intelligent vehicles. and surrounding vehicle strategy ;set up The objective function for the game among surrounding vehicles also depends on the intelligent vehicle strategy. and surrounding vehicle strategy .

[0109] The game objective function of the aforementioned intelligent vehicle includes a lane-merging efficiency game objective function, a right-of-way attribution game objective function, and a driving safety game objective function; the game objective function of surrounding vehicles includes a driving safety game objective function, a right-of-way attribution game objective function, and a driving disturbance game objective function.

[0110] In this invention, the lane-changing efficiency target of intelligent vehicles is measured by lane-changing intrusion, which is defined as the strategy adopted by the intelligent vehicle. The lateral distance of merging into the target lane and the ratio of the lane merging intrusion to the distance between the current lane centerline and the target lane centerline constitute the objective function of the lane merging efficiency game for intelligent vehicles. This invention establishes it as follows:

[0111]

[0112] in, This represents the single-stage duration of a sequential game. This represents the Y-coordinate of the target lane centerline. This indicates the Y-coordinate of the current lane centerline, such as... Figure 2 As shown.

[0113] The aforementioned right-of-way allocation objective for intelligent vehicles is measured by whether the intelligent vehicle acquires road priority. When the intelligent vehicle adopts a strategy... When the intelligent vehicle crosses the lane divider, it gains road priority, and at this point, the payoff of the intelligent vehicle's game objective function is set to 1; when the intelligent vehicle adopts a strategy... However, before crossing the lane divider, the right-of-way still belongs to the surrounding vehicles in the target lane. At this time, the intelligent vehicle does not suffer performance loss, and the objective function payoff of the game is set to 0. This invention establishes the objective function of the right-of-way attribution game for intelligent vehicles as follows:

[0114]

[0115] in, Indicating that smart cars adopt strategies And crosses the lane divider line, Indicating that smart cars adopt strategies However, it has not yet crossed the lane divider line.

[0116] The aforementioned driving safety objectives of intelligent vehicles are achieved by the intelligent vehicle and surrounding vehicles each adopting strategies from their respective policy sets. , The spatial distance between the induced future coordinate positions is measured by a logarithmic function. This invention assumes that the intelligent vehicle only undergoes variable speed motion in the Y direction and uniform speed motion in the X direction. The future spatial distance is positively correlated with the objective function value of the driving safety game. This invention establishes the objective function of the intelligent vehicle's driving safety game as follows:

[0117]

[0118] in, This represents the initial coordinates of the intelligent vehicle. Indicates the initial coordinates of surrounding vehicles. This indicates the longitudinal speed of the intelligent vehicle. This indicates the longitudinal speed of surrounding vehicles.

[0119] The aforementioned driving safety objectives for surrounding vehicles are also determined by the strategies adopted by both surrounding vehicles and the intelligent vehicle from their respective strategy sets. , The objective function induced is equivalent to the driving safety game objective function of intelligent vehicles. This invention establishes the driving safety game objective function of surrounding vehicles as follows:

[0120]

[0121] in, The objective function for the game theory of driving safety of intelligent vehicles.

[0122] The aforementioned right-of-way attribution objective for surrounding vehicles is measured by whether surrounding vehicles continue to have priority on the road. When intelligent vehicles adopt strategies... When a vehicle crosses a lane divider, surrounding vehicles lose their right-of-way. At this point, the objective function payoff for the surrounding vehicles is set to... When intelligent vehicles adopt strategies However, before crossing the lane divider, road priority still belongs to the surrounding vehicles. At this time, the surrounding vehicles have not improved their performance, and the objective function payoff of the game is set to 0. This invention establishes the objective function of the right-of-way attribution game for surrounding vehicles as follows:

[0123]

[0124] in, Indicating that smart cars adopt strategies And crosses the lane divider line, Indicating that smart cars adopt strategies However, it has not yet crossed the lane divider line.

[0125] The aforementioned interference from surrounding vehicles is measured by the relationship between the timeliness of interference and the ideal timeliness of surrounding vehicles. This invention defines the timeliness of interference as the timeliness of surrounding vehicles when they adopt a strategy. The time required for surrounding vehicles to reach the current longitudinal position of the intelligent vehicle is defined as the ideal timeliness, which is the time required for them to reach the current longitudinal position of the intelligent vehicle while maintaining their original speed without engaging in game-like decision-making with the intelligent vehicle. The aforementioned interference includes both promoting and suppressing interference. When surrounding vehicles adopt strategies within the acceleration range from their strategy set to cope with the intelligent vehicle's lateral lane-changing behavior, this is considered promoting interference, and the timeliness of this interference is less than the ideal timeliness. When surrounding vehicles adopt strategies within the deceleration range from their strategy set to cope with the intelligent vehicle's lateral lane-changing behavior, this is considered suppressing interference, and the timeliness of this interference is greater than the ideal timeliness. Based on this, the objective function of the game-like interference of surrounding vehicles is established as follows:

[0126]

[0127]

[0128]

[0129]

[0130] in, This indicates a desire to improve the timeliness of interference. This indicates the timeliness of suppressing interference. It indicates ideal timeliness.

[0131] The objective function of the intelligent vehicle and the objective function of the surrounding vehicles both depend on the strategies adopted by the intelligent vehicle. And the strategy of a certain type of surrounding vehicle in the subjective perception of an intelligent vehicle These are common independent variables. The objective function of the intelligent vehicle game consists of lane-merging efficiency objectives, right-of-way attribution objectives, driving safety objectives, and a type weight factor for the intelligent vehicle; the objective function of the surrounding vehicles game consists of driving safety objectives, right-of-way attribution objectives, driving interference objectives, and a type weight factor for a certain type of surrounding vehicles in the intelligent vehicle's subjective perception; where the type of intelligent vehicle is denoted as... , ∈{radical, moderate, conservative}, with the following weighting factors: The types of surrounding vehicles perceived by intelligent vehicles are: , ∈{radical, moderate, conservative}, with the following weighting factors:

[0132]

[0133] The expression is: if it is an aggressive type, then the weighting factor is the first set. If it is of the moderate type, then the weighting factor is the one in the second set. If it is a conservative approach, then the weighting factor is from the third set. Therefore, the weight factor is a subset of the three sets mentioned above.

[0134] The virtual sequential game model designed in this invention is as follows:

[0135]

[0136]

[0137] in, This represents a virtual game model for intelligent vehicles. This represents a virtual game model of surrounding vehicles in the subjective perception of intelligent vehicles.

[0138] In step S105, the problem of solving the virtual sequential game model is transformed into a model optimization problem:

[0139]

[0140]

[0141]

[0142] in, , These represent the strategies predicted by the intelligent vehicle for surrounding vehicles. , The corresponding strategies are as follows. Game model optimization is achieved through reverse traversal and pruning of the sequential game unfolding tree, outputting the strategy pairs under the game equilibrium state reached by the intelligent vehicle and its subjective perception of surrounding vehicles. This completes the virtual game system for subjective decision-making in intelligent vehicles, and the intelligent vehicles will actually make decisions based on this system. .

[0143] 2. Step S2 establishes an objective decision-making reality game system. This reality game system is the core of the surrounding vehicles' autonomous decision-making. This reality game system shares the definition of key elements of sequential game theory in this invention with the aforementioned virtual game system. The difference is that the surrounding vehicles possess knowledge that the intelligent vehicle's perception of them is subjective, limited, and uncertain, and do not follow the response strategies that the intelligent vehicle believes they will take. Instead of taking concrete action, it should be based on the actual type of the surrounding vehicles. To implement decision-making strategies In order to make the best practical response, ∈{radical, moderate, conservative}, with weighting factors of The expression is: if it is an aggressive type, then the weighting factor is the first set. If it is of the moderate type, then the weighting factor is the one in the second set. If it is a conservative approach, then the weighting factor is from the third set. Therefore, the three sets mentioned above are combined into a union, and the weight factors belong to this union. The real-world sequential game model designed in this invention is as follows:

[0144]

[0145]

[0146] in, This represents a real-world game theory model for intelligent vehicles. This represents a real-world game model involving surrounding vehicles.

[0147] The problem of solving the real-world sequential game model is then transformed into a model optimization problem:

[0148]

[0149] Among them, intelligent vehicles, which occupy the position of first mover in the sequential game, have already made practical decisions. Game theory model optimization only requires optimizing the actual decision-making strategies already made by the intelligent vehicle. As parameters, any response strategy that a real-type surrounding vehicle might take from its policy set. This is an optimization problem with variables. Output the actual optimal response strategy for the surrounding vehicles of the true type. And based on this, make the best actual response decision.

[0150] 3. In step S3, strategy verification is involved, and the intelligent vehicle obtains the actual optimal response strategy of surrounding vehicles. This confirms the optimal response strategy adopted by surrounding vehicles as perceived in the virtual game system. The resulting deviation between the payoffs of the self-game objective function:

[0151]

[0152] in, This represents the deviation of the objective function payoff of intelligent vehicles in virtual and real-world game systems.

[0153] 4. In step S4, the understanding is refreshed, and the intelligent vehicle determines the payoff deviation caused by the actual optimal response strategy of the surrounding vehicles and the optimal response strategy adopted in the virtual game system's cognition. Correcting one's understanding of the types of vehicles around them, refreshing one's beliefs about intelligent vehicles, and forming a new understanding of the types of vehicles around them. .

[0154] 5. Finally, in the next stage of the parallel operation of the virtual and real game systems, the intelligent vehicle reduces the cognitive gain bias, accurately estimates the type of surrounding vehicles based on the new cognition formed, and makes more reasonable human-like decisions. It repeats all the above steps S1 to S4 until the intelligent vehicle achieves cognitive convergence in the virtual game system of subjective decision-making and the real game system of objective decision-making, and completes the driving task of successfully merging into the target lane.

[0155] Based on the same inventive concept, this invention also provides an intelligent vehicle human-like decision-making system based on a parallel virtual-real game theory dual system. Since the principle by which this system solves the problem is similar to the aforementioned intelligent vehicle human-like decision-making method based on a parallel virtual-real game theory dual system, the implementation of this system can refer to the implementation of the aforementioned method; repeated details will not be elaborated further. (Refer to...) Figure 4 As shown, it specifically includes:

[0156] The subjective decision-making module is used by intelligent vehicles to establish a virtual game system with surrounding vehicles based on their own cognitive knowledge and understanding of the types of surrounding vehicles.

[0157] The objective decision-making module is used to establish a realistic game system between the intelligent vehicle and surrounding vehicles after the surrounding vehicles have made actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle.

[0158] The strategy verification module is used by intelligent vehicles to obtain the actual optimal response strategies of surrounding vehicles and verify the deviation between these strategies and the optimal response strategies perceived by the virtual game system.

[0159] The understanding refresh module is used by intelligent vehicles to correct their understanding of the types of surrounding vehicles based on deviations and refresh their beliefs to form a new cognition of the types of surrounding vehicles.

[0160] In the next stage of parallel operation of the dual systems, the above modules are repeatedly executed, and the intelligent vehicle makes corresponding human-like decisions based on the new cognition until the cognition of the dual systems converges and the intelligent vehicle achieves its lane-merging goal; the dual systems consist of a virtual game system and a real game system.

[0161] In the subjective decision-making module, the intelligent vehicle establishes a virtual game system with surrounding vehicles based on its own cognitive knowledge and understanding of the types of surrounding vehicles, referring to... Figure 5 As shown, it includes a type estimation module, a sequential game allocation module, an action prediction module, a virtual sequential game model design module, and a game model optimization module.

[0162] The type estimation module enables the intelligent vehicle to estimate the type of surrounding vehicles based on the lateral and longitudinal driving behaviors and driving information of surrounding vehicles in the target lane during the recorded historical interaction time domain; the types include: aggressive, moderate and conservative.

[0163] The sequential game allocation module, according to the sequential game rules, positions the intelligent vehicle itself as the forerunner in the sequential game and the surrounding vehicles as the follower in the sequential game.

[0164] The action prediction module, based on the intelligent vehicle's estimation of the types of surrounding vehicles and the sequential game action allocation, predicts the response strategies of surrounding vehicles one by one for all the optional strategies in the intelligent vehicle's own game strategy set.

[0165] The virtual sequential game model design module constructs a virtual sequential game model based on the game objective function of the intelligent vehicle and the game objective function of the surrounding vehicles, as well as based on the type of intelligent vehicle and the type of estimation of the surrounding vehicles.

[0166] in, This represents a virtual game model for intelligent vehicles; the type of intelligent vehicle is denoted as... , ∈{radical, moderate, conservative}, corresponding weight factors are ;

[0167] This represents a virtual game model of surrounding vehicles in the subjective perception of an intelligent vehicle; the types of surrounding vehicles in the subjective perception of an intelligent vehicle are: , ∈{radical, moderate, conservative}, with weighting factors of .

[0168] The game model optimization module transforms the problem of solving the virtual sequential game model into a model optimization problem. It achieves the optimization solution of the virtual sequential game model by reverse traversal and pruning of the sequential game expansion tree, and outputs the strategy pairs under the game equilibrium state reached by the intelligent vehicle and the surrounding vehicles in the intelligent vehicle's subjective perception, so that the intelligent vehicle can make actual decisions.

[0169] In the objective decision-making module, surrounding vehicles make actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, establishing a realistic game system with the intelligent vehicle, referencing... Figure 6 As shown, it includes a real-world sequential game model design module and a game model optimization module; among which, Figure 6 The system also includes a real-world observation module, which means that after the intelligent car, which is the first player in the sequential game, makes a real-world decision in the virtual game system that conforms to the equilibrium state of the game, the surrounding vehicles, which are the followers in the sequential game, observe the real-world strategy adopted by the intelligent car and then engage in a real-world game with the intelligent car.

[0170] The real-world sequential game model design module involves surrounding vehicles making actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, and then the surrounding vehicles responding according to their own true type. To implement decision-making strategies To make the optimal response in practice; the intelligent vehicle establishes a realistic game model with surrounding vehicles as follows:

[0171]

[0172]

[0173] in, This represents a real-world game theory model for intelligent vehicles. ∈{radical, moderate, conservative}, corresponding weight factors are ; This represents the actual decisions made by the intelligent vehicle based on the virtual game system. Represents a real-world game model involving surrounding vehicles;

[0174] Game model optimization module: Transforms the real-world game model solution problem into a model optimization problem.

[0175]

[0176] Among them, intelligent vehicles, which occupy the position of first mover in the sequential game, have already made practical decisions. It only requires optimizing the actual decision-making strategies already made by smart cars. As parameters, any response strategy that a real-type surrounding vehicle might take from its policy set. This is an optimization problem with variables; the output should be the actual optimal response strategy for the surrounding vehicles of the true type. And based on this, the surrounding vehicles make the best actual response decisions.

[0177] In the strategy verification module, the intelligent vehicle obtains the actual optimal response strategies of surrounding vehicles. This confirms the optimal response strategy adopted by surrounding vehicles as perceived in the virtual game system. The resulting deviation between the payoffs of the self-game objective function:

[0178]

[0179] in, This represents the deviation of the objective function payoff of intelligent vehicles in virtual and real-world game systems.

[0180] In the understanding refresh module, the intelligent vehicle determines the payoff deviation caused by the difference between the actual optimal response strategy of surrounding vehicles verified by the strategy verification module and the optimal response strategy adopted in the perception of the virtual game system. Correcting one's understanding of the types of vehicles around them, refreshing one's beliefs about intelligent vehicles, and forming a new understanding of the types of vehicles around them. .

[0181] This invention provides an intelligent vehicle human-like decision-making system based on a parallel virtual-real game theory dual system. Through the parallel integrated operation of these two systems, it achieves bidirectional cognitive collaboration and information exchange. The virtual and real game theory systems interact, continuously correcting and updating the intelligent vehicle's perception of surrounding vehicles, ensuring consistency with actual observations and optimal responses. This parallel integrated operation mode can better improve the intelligent vehicle's decision-making capabilities and response speed, making it more adaptable to complex traffic environments.

[0182] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0183] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A human-like decision-making method for intelligent vehicles based on a dual-system parallel virtual-real game theory approach, characterized in that: Includes the following steps: S1. Intelligent vehicles, based on their own cognitive knowledge and understanding of surrounding vehicle types, establish a virtual game system with surrounding vehicles; including: S101. The intelligent vehicle estimates the type of surrounding vehicles based on the lateral and longitudinal driving behaviors and driving information of surrounding vehicles in the target lane during the recorded historical interaction time domain; the types include: aggressive, moderate and conservative. S102. According to the sequential game rules, the intelligent vehicle itself is positioned as the forerunner in the sequential game, and the surrounding vehicles are positioned as the follower in the sequential game. S103. Based on the intelligent vehicle's estimation of the types of surrounding vehicles and the sequential game action allocation, predict the response strategies of surrounding vehicles one by one for all the optional strategies in the intelligent vehicle's own game strategy set. S104. Based on the game objective function of the intelligent vehicle and the game objective function of the surrounding vehicles, and based on the type of intelligent vehicle and the type of estimation of the surrounding vehicles, construct a virtual sequential game model. S105. The problem of solving the virtual sequential game model is transformed into a model optimization problem. The virtual sequential game model is optimized and solved by reverse traversal and pruning of the sequential game expansion tree. The strategy pair under the game equilibrium state reached by the intelligent car and the surrounding vehicles in the subjective perception of the intelligent car is output so that the intelligent car can make actual decisions. S2. After the surrounding vehicles make actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, the surrounding vehicles establish a real-world game system with the intelligent vehicle. S3. The intelligent vehicle obtains the actual optimal response strategy of the surrounding vehicles and verifies the deviation between the strategy and the optimal response strategy of the surrounding vehicles as perceived in the virtual game system. The actual optimal response strategy of the surrounding vehicles is obtained through the following process: the surrounding vehicles make decisions based on their true type, the intelligent vehicle establishes a real game model with the surrounding vehicles and optimizes the solution, and outputs the actual optimal response strategy of the surrounding vehicles of the true type. S4. Intelligent vehicles correct their understanding of the types of surrounding vehicles based on deviations and refresh their beliefs, forming a new perception of the types of surrounding vehicles. S5. In the next stage of parallel operation of the dual systems, repeat steps S1-S4 above. The intelligent vehicle makes corresponding human-like decisions based on the new cognition until the cognition of the dual systems converges and the intelligent vehicle achieves its lane-merging goal. The dual systems consist of a virtual game system and a real game system.

2. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 1, characterized in that, The process of constructing the game objective function of the intelligent vehicle in step S104 is as follows: Based on the rules of sequential game theory, and considering the arbitrary lateral acceleration strategies that intelligent vehicles can choose... And any longitudinal acceleration strategy selectable by surrounding vehicles Constructing the game objective function for intelligent vehicles ; Among them, the game objective function of intelligent vehicles This includes: the objective function of lane merging efficiency game, the objective function of right-of-way attribution game, and the objective function of driving safety game; 1) The objective function of the parallel track efficiency game is as follows: in, This represents the single-stage duration of a sequential game. This represents the Y-coordinate of the target lane centerline. This indicates the Y-coordinate of the current lane centerline; 2) The objective function of the right-of-way attribution game is as follows: in, Indicating that smart cars adopt strategies It crosses the lane divider, gains road priority, and sets the payoff of the intelligent vehicle's right-of-way attribution game objective function to 1; Indicating that smart cars adopt strategies However, since the lane divider has not yet been crossed, the right of way still belongs to the surrounding vehicles in the target lane, so the payoff of the intelligent vehicle's right-of-way attribution game objective function is set to 0. 3) The objective function for the driving safety game is as follows: in, This represents the initial coordinates of the intelligent vehicle. Indicates the initial coordinates of surrounding vehicles. This indicates the longitudinal speed of the intelligent vehicle. This indicates the longitudinal speed of surrounding vehicles.

3. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 2, characterized in that, The objective function for the game between surrounding vehicles in step S104 is constructed as follows: Based on the rules of sequential game theory, and considering the arbitrary lateral acceleration strategies that intelligent vehicles can choose... And any longitudinal acceleration strategy selectable by surrounding vehicles Construct the game objective function of the surrounding vehicles. ; Among them, the objective function of the surrounding vehicles in the game. This includes: the objective function of driving safety game, the objective function of right-of-way attribution game, and the objective function of driving disturbance game; a) The objective function for the driving safety of surrounding vehicles is as follows: The driving safety objective for surrounding vehicles is determined by the strategies adopted by both surrounding vehicles and the intelligent vehicle from their respective policy sets. , The objective function of the game of driving safety of surrounding vehicles induced by this. Equivalent to the objective function of the driving safety game for intelligent vehicles ; b) The objective function of the right-of-way attribution game for surrounding vehicles is as follows: in, Indicating that smart cars adopt strategies And cross the lane dividing line, surrounding vehicles lose their right-of-way, and the objective function payoff of the right-of-way attribution game for surrounding vehicles is set to -1; Indicating that smart cars adopt strategies However, since the lane divider has not been crossed, the surrounding vehicles have not improved their performance, so the payoff of the right-of-way attribution game objective function for the surrounding vehicles is set to 0. c) The objective function for the game of interference from surrounding vehicles is as follows: The driving interference of surrounding vehicles is measured by the relationship between the interference timeliness and the ideal timeliness of surrounding vehicles; interference timeliness is defined as the timeliness of surrounding vehicles in taking strategies. The time required to reach the current longitudinal position of the intelligent vehicle; ideal timeliness is defined as the time required for surrounding vehicles to reach the current longitudinal position of the intelligent vehicle while maintaining their original speed without engaging in game-like interactive decision-making with the intelligent vehicle; the driving interference target includes promoting interference and suppressing interference; in, The objective function represents the game objective function for dealing with the interference of surrounding vehicles. This indicates that when surrounding vehicles adopt strategies within the acceleration range from their strategy set in order to cope with the lateral lane-changing behavior of the intelligent vehicle, it is considered to promote interference timeliness, which is less than ideal timeliness. This indicates that when surrounding vehicles adopt a strategy within the deceleration range from their strategy set in order to cope with the lateral lane-changing behavior of the intelligent vehicle, it belongs to the timeliness of interference suppression, which is greater than the ideal timeliness. It indicates ideal timeliness.

4. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 3, characterized in that, In step S104, the virtual sequential game model is constructed as follows: in, This represents a virtual game model for intelligent vehicles; the type of intelligent vehicle is denoted as... , ∈{radical, moderate, conservative}, corresponding weight factors are ; This represents a virtual game model of surrounding vehicles in the subjective perception of an intelligent vehicle; the types of surrounding vehicles in the subjective perception of an intelligent vehicle are: , The weighting factor is .

5. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 4, characterized in that, In step S105, the problem of solving the virtual sequential game model is transformed into a model optimization problem, including: in, , These represent the strategies predicted by the intelligent vehicle for surrounding vehicles. , Response strategies; This represents the set of strategies for surrounding vehicles. For elements in the policy set; This represents the strategy pair in the game equilibrium state reached by the intelligent vehicle in its output and the surrounding vehicles in the intelligent vehicle's subjective perception.

6. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 5, characterized in that, Step S2 includes: S201. After the surrounding vehicles have made actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle, the surrounding vehicles then respond according to their own actual types. To implement decision-making strategies In order to make the best practical response; S202. The intelligent vehicle establishes a realistic game model with surrounding vehicles as follows: in, This represents a real-world game theory model for intelligent vehicles. ∈{radical, moderate, conservative}, corresponding weight factors are ; This represents the actual decisions made by the intelligent vehicle based on the virtual game system. Represents a real-world game model involving surrounding vehicles; S203. Transform the problem of solving the real-world game model into a model optimization problem: Among them, intelligent vehicles, which occupy the position of first mover in the sequential game, have already made practical decisions. It only requires optimizing the actual decision-making strategies already made by smart cars. As parameters, any response strategy that a real-type surrounding vehicle might take from its policy set. This is an optimization problem with variables; the output should be the actual optimal response strategy for the surrounding vehicles of the true type. And based on this, the surrounding vehicles make the best actual response decisions.

7. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 6, characterized in that, Step S3 includes: Intelligent vehicles learn the actual optimal response strategies of surrounding vehicles. This confirms the optimal response strategy adopted by surrounding vehicles as perceived in the virtual game system. The resulting deviation between the payoffs of the self-game objective function: in, This represents the deviation of the objective function payoff of intelligent vehicles in virtual and real-world game systems.

8. The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game theory according to claim 7, characterized in that, The payoff bias caused by the intelligent vehicle's response strategy based on the actual optimal response strategies of surrounding vehicles and the optimal response strategies adopted in the virtual game system's perception. Correcting one's understanding of the types of vehicles around them, refreshing one's beliefs about intelligent vehicles, and forming a new understanding of the types of vehicles around them. .

9. A human-like decision-making system for intelligent vehicles based on a dual-system parallel virtual-real game theory approach, characterized in that: The intelligent vehicle human-like decision-making method based on a dual-system parallel virtual-real game as described in any one of claims 1-8, wherein the system comprises: The subjective decision-making module is used by intelligent vehicles to establish a virtual game system with surrounding vehicles based on their own cognitive knowledge and understanding of the types of surrounding vehicles. The objective decision-making module is used to establish a realistic game system between the surrounding vehicles and the intelligent vehicle after the surrounding vehicles have made actual observations and optimal responses based on the strategies already adopted by the intelligent vehicle. The strategy verification module is used by intelligent vehicles to obtain the actual optimal response strategies of surrounding vehicles and verify the deviation between these strategies and the optimal response strategies perceived by the virtual game system. The understanding refresh module is used by intelligent vehicles to correct their understanding of the types of surrounding vehicles based on deviations and refresh their beliefs to form a new cognition of the types of surrounding vehicles. In the next stage of parallel operation of the dual systems, the above modules are repeatedly executed, and the intelligent vehicle makes corresponding human-like decisions based on the new cognition until the cognition of the dual systems converges and the intelligent vehicle achieves its lane-merging goal; the dual systems consist of a virtual game system and a real game system.

Images