Autonomous Vehicle Lane Changing Behavior Decision-Making Method in Roundabout Scenarios Based on Hybrid Game Theory

Abstract

This invention proposes a hybrid game theory-based decision-making method for lane-changing behavior of autonomous vehicles in roundabout scenarios, comprising the following steps: First, determining the driving style; second, designing rules, assuming that the route of the vehicle after entering the roundabout is clear under ideal conditions. Game theory decision-making between vehicles is conducted by defining three concepts: "entry guidance zone," "exit guidance zone," and "buffer zone" of the roundabout. Using a hybrid game theory approach, the value reward of the vehicle in future driving segments is calculated by comprehensively considering the probability of style and driving intention, resulting in the driving behavior at the next moment. Finally, the calculated driving behavior is applied to the vehicle. This invention solves the problems of difficult lane-changing decisions, risks, and low efficiency caused by the lack of research in the roundabout domain and the difficulty in accurately judging driving style in existing technologies.

Description

Technical Field

[0001] This invention belongs to the field of autonomous vehicle simulation testing, and in particular relates to a lane-changing behavior decision-making method for autonomous vehicles in a roundabout scenario based on hybrid game theory. Background Technology

[0002] This study analyzes the potential scenarios and behavioral decisions of vehicles navigating roundabouts. Data shows that 20% of traffic fatalities are related to road intersections. Roundabouts are variable and interactive scenarios; by eliminating traffic lights and stop points, dynamic traffic flow can be improved, delays and congestion reduced. This is crucial for enhancing the safety, traffic efficiency, and energy conservation and emission reduction of autonomous vehicles, and also provides a method for autonomous vehicles to collect data on new road conditions.

[0003] Game theory can be used to optimize traffic light control strategies at roundabout intersections. A game-theoretic relationship exists between autonomous vehicles and intersection signals; game theory models can be used to intelligently optimize these signals, maximizing the overall efficiency of the intersection. The application of game theory can promote the intelligence and collaboration of autonomous driving systems in roundabout scenarios, improving overall intersection traffic efficiency, reducing congestion, and increasing intersection safety. By comprehensively considering various factors and the interests of all participants, game theory provides a powerful tool for autonomous driving systems, enabling them to better adapt to complex traffic environments.

[0004] In summary, roundabouts are among the most congested and complex intersections in urban traffic. In roundabout scenarios, vehicles need to make lane-changing decisions based on traffic flow and their own driving needs to ensure traffic safety and efficiency.

[0005] However, due to the mutual influence and competition between vehicles, the lane-changing decision-making problem becomes complex and difficult. Therefore, researching lane-changing decision-making methods for roundabout scenarios based on hybrid game theory is of significant background and importance. Summary of the Invention

[0006] The purpose of this invention is to provide a lane-changing behavior decision-making method for autonomous vehicles in roundabout scenarios based on hybrid game theory, solving the problems of difficult decision-making, risks, and low efficiency caused by the lack of research in roundabout scenarios and the inability to accurately judge driving styles in existing technologies. To achieve the above objective, the technical solution adopted is as follows:

[0007] A method for determining lane-changing behavior of autonomous vehicles in a roundabout scenario based on hybrid game theory includes the following steps:

[0008] Step 1: Under the premise of vehicle-to-everything (V2X) connectivity, capture driver behavior data and perform fuzzy inference to determine their driving style;

[0009] Step 2: Design soft and hard rules suitable for roundabout traffic simulation;

[0010] Step 3: Divide the vehicle operation area between adjacent traffic intersections of the roundabout into four categories: entrance guidance area, buffer zone I, buffer zone II, and exit guidance area. Based on soft and hard rules, establish the corresponding vehicle behavior for each type of vehicle operation area.

[0011] Step 4: Based on vehicle behavior and corresponding decision-making mechanisms, and combined with the collected vehicle status or distance, determine whether each type of vehicle behavior requires game theory or the definition of a game alliance. If so, proceed to step 5.

[0012] Step 5: Receive the vehicle states output by all vehicle dynamics models in Step 4, combine them with driving style, and output the vehicle states after the game, that is, the vehicle states of each vehicle dynamics model under the optimal solution.

[0013] Step 6: The vehicle dynamics model executes the vehicle state under the optimal solution.

[0014] Preferably, the soft and hard rules in step 2 specifically include:

[0015] Rule 1: When exiting from an adjacent intersection, travel only in the outermost lane;

[0016] Rule 2: When exiting from a position one intersection away, you must first enter the middle lane, then move to the outermost lane, and finally exit.

[0017] Rule 3: When exiting from a position two intersections apart, you must first enter the middle lane, then move to the outermost lane, and finally exit.

[0018] Rule 4: When exiting from a position three intersections apart, you must first enter the middle lane, then the innermost lane, and then exit outwards in sequence, finally leaving the exit.

[0019] Rule 5: When exiting from the entrance, you must first enter the middle lane, then the innermost lane, then enter the middle lane again, pass through two forks, then move to the outermost lane, and finally leave the exit.

[0020] Preferably, step 3 specifically includes the following steps:

[0021] Step 3A: Divide the vehicle operation area between adjacent traffic intersections of the roundabout into four categories: entrance guidance area, buffer zone I, buffer zone II, and exit guidance area;

[0022] Step 3B: Establish vehicle behavior corresponding to the operating area of ​​each type of vehicle.

[0023] Vehicle behavior one and vehicle behavior two occurred in the entrance guidance area;

[0024] Among them, vehicle behavior one is a car entering the outer lane from entrance A; vehicle behavior two is a car entering the middle lane from the inner lane.

[0025] Vehicle behavior three and vehicle behavior four occur in buffer zone I;

[0026] Vehicle behavior three: G-class vehicles in the outer lane will enter the middle lane; vehicle behavior four: R-class vehicles in the middle lane will enter the inner lane.

[0027] Vehicle behavior five occurred in buffer zone II, which was a Class P vehicle in the middle lane entering the outer lane.

[0028] Vehicle behavior six will occur in the exit guidance area, which is a vehicle in the outer lane leaving Exit B.

[0029] Preferably, the decision-making mechanism in step 4 is as follows:

[0030] Decision-making mechanism one, corresponding to vehicle behavior one, is as follows:

[0031] When 2.5≤d2-d1≤5, the two cars cooperate and play a game to enter the entrance.

[0032] When d2-d1 is not within the above range, vehicles with smaller d1 and d2 values ​​enter the roundabout first;

[0033] Decision-making mechanism two, corresponding to vehicle behavior two, is as follows:

[0034] In the middle lane, if a car from the inner lane enters the middle lane and a Class G car enters the middle lane, is there any other vehicle within 5 meters? If so, the car and the other vehicles within 5 meters engage in a cooperative game.

[0035] Decision-making mechanism three applies to vehicle behavior three and vehicle behavior four, specifically as follows:

[0036] For a Class G vehicle in vehicle behavior three, if there is a vehicle of the same type behind it and the distance between the two vehicles is less than 3 meters, then the two Class G vehicles are defined as a sub-alliance. Then, the Class R vehicles and the sub-alliance in vehicle behavior four are defined as a large alliance.

[0037] When a Class G vehicle enters Buffer Zone I in Vehicle Behavior 3, the vehicles within 3 meters behind the Class G vehicle, the vehicles in the middle lane parallel to the front of the Class G vehicle, and the vehicles in the outer lane within 5 meters behind the Class G vehicle and traveling in the middle lane, together form a coalition game.

[0038] Decision-making mechanism four corresponds to vehicle behavior five, specifically as follows:

[0039] The P-class vehicles in the middle lane, the vehicles in the outer lane that are parallel to the front of the P-class vehicles in the middle lane, and the vehicles in the outer lane that are within 5 meters behind the P-class vehicles in the middle lane constitute the alliance game in decision-making mechanism four.

[0040] Decision-making mechanism five corresponds to vehicle behavior six. Specifically, in the exit guidance area, if there is a vehicle within 3 meters of lane Bo-1, then enter lane Bo-2; otherwise, enter lane Bo-1.

[0041] Preferably, step 1 specifically includes the following steps:

[0042] Step 1A: Collect driver behavior data, including: emergency braking, acceleration, turning, and vehicle speed;

[0043] Step 1B: Based on fuzzy inference, classify the collected acceleration and vehicle speed using fuzzy sets;

[0044] Step 1C: Combining the fuzzy set from Step 1B, the data collected in Step 1A is transformed into driving style.

[0045] An autonomous driving testing system includes:

[0046] Traffic simulator used to generate roundabout road networks;

[0047] An interactive game-theoretic decision-maker, designed for simulation needs, takes the vehicle speed output from the vehicle dynamics model as input and embeds a game-theoretic decision-making algorithm based on logical rules. Considering the scenario of vehicles running in a roundabout, it divides the vehicle running area between adjacent intersections of the roundabout into four categories: entrance guidance area, buffer area I, buffer area II, and exit guidance area. It constructs a decision-making mechanism for each type of running area, determines whether the running area needs to engage in a game or define a game alliance, and if so, outputs the vehicle state after the game.

[0048] An interactive game-theoretic controller is used to output control quantities for autonomous driving based on the vehicle state output by the interactive game-theoretic decision-maker.

[0049] And a vehicle dynamics model, used to simulate the real response of the vehicle's mechanical system to the output of the interactive game controller, inputting throttle opening, braking pressure and front wheel deflection angle control quantities, and outputting the real vehicle state;

[0050] The vehicle status includes: speed, acceleration, and driving trajectory.

[0051] Compared with the prior art, the advantages of the present invention are:

[0052] In a roundabout scenario, when a vehicle decides its lane-changing probability, it considers multiple scenarios based on the payoff matrix formed by the vehicle and surrounding vehicles of different styles. It then uses Nash equilibrium to solve for the global optimum in the cost function of each vehicle, outputting the motion state of each vehicle under the optimal solution. Finally, the data is transmitted back to the vehicle, and the game is repeated in the next time step to find the optimal solution. Attached Figure Description

[0053] Figures 1-5 This is a diagram illustrating both hard and soft rules;

[0054] Figure 6 This is a random example diagram showing the traffic flow between intersections A and B in a roundabout.

[0055] Figures 7-8 This is a magnified view of a portion of the entrance guidance area;

[0056] Figures 9-10 This is a magnified view of a portion of buffer zone I;

[0057] Figure 11 This is a magnified view of a portion of buffer zone II;

[0058] Figure 12 This is a magnified view of a portion of the export guidance area. Detailed Implementation

[0059] The following will describe in more detail the lane-changing behavior decision-making method for autonomous vehicles in a roundabout scenario based on hybrid game theory, with reference to the schematic diagrams. Preferred embodiments of the invention are illustrated. It should be understood that those skilled in the art can modify the invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the invention.

[0060] A lane-changing behavior decision-making method for autonomous vehicles in a roundabout scenario based on hybrid game theory is proposed, based on an autonomous driving test system, which includes:

[0061] Traffic simulator used to generate roundabout road networks.

[0062] An interactive game-theoretic decision-maker, designed for simulation needs, takes the vehicle speed output from the vehicle dynamics model as input and embeds a game-theoretic decision-making algorithm based on logical rules. Considering the scenario of vehicles running in a roundabout, it divides the vehicle running area between adjacent intersections of the roundabout into four categories: entrance guidance area, buffer area I, buffer area II, and exit guidance area. It constructs a decision-making mechanism for each type of running area, determines whether the running area needs to engage in a game or define a game alliance, and if so, outputs the vehicle state after the game.

[0063] An interactive game-theoretic controller is used to output control quantities for autonomous driving based on the vehicle state output by the interactive game-theoretic decision-maker.

[0064] The vehicle dynamics model is used to simulate the real response of the vehicle's mechanical system to the output of the interactive game-like controller. The inputs are throttle opening, braking pressure and front wheel deflection angle control quantities, and the output is the real vehicle state.

[0065] The vehicle status includes: speed, acceleration, and driving trajectory.

[0066] like Figures 1-12 The decision-making method for lane-changing behavior of autonomous vehicles in a roundabout scenario based on hybrid game theory includes the following steps:

[0067] Step 1: Under the premise of vehicle-to-everything (V2X) connectivity, capture driver behavior data and perform fuzzy inference to determine their driving style.

[0068] Step 1A: Collect driver behavior data through vehicle sensors (such as accelerometers, speedometers, etc.), including: emergency braking, acceleration, turning and vehicle speed.

[0069] Emergency braking data typically refers to parameters generated during sudden, emergency braking of a vehicle. This data can be acquired through various sensors. Specifically, the following technologies can help identify emergency braking situations:

[0070] Speed ​​sensor: The speed sensor on a vehicle can monitor changes in the vehicle's speed. When the speed drops rapidly, it indicates that the vehicle is undergoing emergency braking.

[0071] Accelerometer sensors: These sensors measure a vehicle's acceleration or deceleration. During emergency braking, a vehicle experiences significant deceleration, which the acceleration sensors can detect.

[0072] Braking system monitoring: Vehicle braking systems usually have monitoring devices that can detect sudden pressure on the brake pedal, thus determining that the driver is performing emergency braking.

[0073] Vehicle dynamic control system: Advanced vehicles may be equipped with vehicle dynamic control systems, such as Electronic Stability Program (ESP). These systems can monitor and process the vehicle's driving status and take measures to help the driver control the vehicle when emergency braking is detected.

[0074] Vehicle-to-everything (V2X) communication: In more advanced technologies, vehicles can exchange data with surrounding vehicles via wireless communication technology. If a vehicle is undergoing emergency braking, it can use this method to promptly notify other vehicles, preventing accidents even in poor visibility conditions.

[0075] Data on a vehicle's turning behavior can be monitored and recorded using various sensors and systems. Here are some commonly used techniques and indicators that can help identify a vehicle's turning behavior:

[0076] Steering angle sensors: These sensors measure the angle of rotation of the steering wheel to determine the degree of turning of the vehicle.

[0077] Vehicle dynamic control systems (such as ESP): Systems such as Electronic Stability Program (ESP) can determine whether a vehicle is turning by analyzing the speed and steering angle of the wheels.

[0078] Gyroscope: A gyroscope is a sensor that can measure the rotational motion of an object. It can be used to detect the turning motion of a vehicle.

[0079] Accelerometer: When a vehicle is turning, the acceleration sensor detects the vehicle's lateral acceleration because the vehicle experiences centripetal force during the turn.

[0080] Wheel speed sensors: The speed sensors on each wheel can provide information on the wheel's rotational speed. By comparing the rotational speeds of different wheels, the vehicle's turning behavior can be determined.

[0081] Lane Keeping Assist System (LKAS): This type of system monitors the vehicle's position in the lane using cameras or radar, and issues a warning when the vehicle begins to turn and deviate from the lane.

[0082] Step 1B: Based on fuzzy inference, fuzzy sets are used to classify the collected acceleration and vehicle speed.

[0083] Fuzzy sets include: fuzzy sets with medium speed, fuzzy sets with low speed, fuzzy sets with high speed, fuzzy sets with fast acceleration, fuzzy sets with slow acceleration, and fuzzy sets with medium acceleration.

[0084] If the vehicle speed is between 15 and 30 km / h, it is classified as a fuzzy set of medium speed.

[0085] If the vehicle speed is below 15 km / h, it is classified as a fuzzy set of low speeds;

[0086] If the vehicle speed is higher than 30 km / h, it is classified as a fuzzy set of higher speeds.

[0087] If the car's acceleration is greater than 1 m / s² 2 It is classified as a fuzzy set of fast acceleration;

[0088] If the vehicle's acceleration is less than 0.5 m / s² 2 It is classified as a fuzzy set with slow acceleration;

[0089] If the car's acceleration is 0.5 m / s² 2 Up to 1m / s 2 Between these, it is classified as a fuzzy set with moderate acceleration.

[0090] Step 1C: Combining the fuzzy set from Step 1B, the data collected in Step 1A is transformed into driving style.

[0091] If the vehicle speed is high (the vehicle speed belongs to the fuzzy set of high speed) and the vehicle acceleration belongs to the fuzzy set of fast acceleration or medium acceleration, then the driver's driving style is aggressive.

[0092] If the vehicle speed is high (the vehicle speed belongs to the fuzzy set of high speeds) and the vehicle acceleration belongs to the fuzzy set of slow accelerations, then the driver's driving style is stable.

[0093] If the vehicle speed is moderate (the vehicle speed belongs to the fuzzy set of medium speeds), then the driver's driving style is stable.

[0094] If the vehicle speed is slow (the vehicle speed belongs to the fuzzy set of low speeds) and the vehicle acceleration belongs to the fuzzy set of fast accelerations, then the driver's driving style is stable.

[0095] If the vehicle speed is slow (the vehicle speed belongs to the fuzzy set of low speeds) and the vehicle acceleration belongs to the fuzzy set of medium or slow accelerations, then the driver's driving style is conservative.

[0096] Step 2: Design soft and hard rules suitable for roundabout traffic simulation.

[0097] Rule 1: When exiting from an adjacent intersection, only travel in the outermost lane. For example... Figure 1 As shown.

[0098] Rule 2: When exiting from a point one intersection away, you must first enter the middle lane, then move to the outermost lane, and finally exit. For example... Figure 2 As shown.

[0099] Rule 3: When exiting from a point two intersections apart, you must first enter the middle lane, then move to the outermost lane, and finally exit. For example... Figure 3 As shown.

[0100] Rule 4: When exiting at a point three intersections apart, you must first enter the middle lane, then the innermost lane, and then exit in sequence, finally leaving the exit. For example... Figure 4 As shown.

[0101] Rule 5: When exiting from the entrance, you must first enter the middle lane, then the innermost lane, then enter the middle lane again, pass through two forks, then move to the outermost lane, and finally exit. For example... Figure 5 As shown.

[0102] Step 3: Divide the vehicle operation area between adjacent traffic intersections of the roundabout into four categories: entrance guidance area, buffer zone I, buffer zone II, and exit guidance area. Based on soft and hard rules, establish the corresponding vehicle behavior for each type of vehicle operation area.

[0103] Step 3A: Divide the vehicle operation area between adjacent traffic intersections of the roundabout into four categories: entrance guidance area, buffer zone I, buffer zone II, and exit guidance area.

[0104] like Figure 6 As shown, the roundabout consists of five intersections: A, B, C, D, and E.

[0105] Each intersection has two lanes for entering and two lanes for exiting.

[0106] The area between the center lines of intersections A and B forms an example area.

[0107] The roundabout consists of five such areas.

[0108] The area is divided into three parts: the entrance guidance area, the buffer zone, and the exit guidance area.

[0109] The buffer zone is further divided into buffer zone I and buffer zone II.

[0110] Category G vehicles are those that ultimately leave the buffer zone via the middle lane;

[0111] R-class vehicles are those that ultimately leave the buffer zone via the inner lane;

[0112] Class P vehicles are those that ultimately leave the buffer zone via the outer lane.

[0113] Among them, such as Figure 6 In the random diagram shown, only vehicles of type P exit from intersection B.

[0114] Step 3B: Establish vehicle behavior corresponding to the operating area of ​​each type of vehicle.

[0115] Two behaviors may occur in the entrance guidance area: a car entering the outer lane from entrance A (vehicle behavior one) and a car entering the middle lane from the inner lane (vehicle behavior two).

[0116] Vehicle Behavior 1: Vehicles at Entrance A first enter the roundabout via the outer lane (e.g., Figure 1 (As shown), then determine whether to stay in the outer lane or enter the middle lane;

[0117] Vehicle Behavior 2: such as Figure 4 , Figure 5 As shown, vehicles only make a brief stop in the inner lane before exiting the entry section. Therefore, all vehicles in the inner lane will exit at the next entrance guidance area.

[0118] Two behaviors may occur in Buffer Zone I: a Class G vehicle in the outer lane may enter the middle lane (Vehicle Behavior 3) and a Class R vehicle in the middle lane may enter the inner lane (Vehicle Behavior 4).

[0119] Vehicle Behavior 3: In the previous section, all vehicles in the outer lanes exited the roundabout at Exit A. Therefore, at this time, the outer lanes only contain vehicles that entered through Exit A. If a vehicle exits through Exit B, it remains in the outer lane (e.g., ...). Figure 1As shown), if a vehicle exits from another exit, it enters the middle lane in this area (e.g., Figure 2-5 (as shown);

[0120] Vehicle Behavior 4: When a vehicle in the middle lane... Figure 4-5 When it is necessary to enter the inner lane, the cars in the inner lane have already entered the middle lane from the entrance guidance area. Therefore, when vehicle behavior occurs, the absence of vehicles in the inner lane will affect the behavior.

[0121] In buffer zone II, a behavior may occur where a Class P vehicle in the middle lane enters the outer lane (vehicle behavior five).

[0122] Vehicle Behavior 5: When a vehicle in the middle lane needs to exit from Exit B, it must at this time enter the outer lane (e.g., Figure 2-5 (As shown).

[0123] In the exit guidance area, a behavior may occur where vehicles in the outer lane leave Exit B (Vehicle Behavior Six).

[0124] Vehicle Behavior 6: As the rules indicate, all vehicles in the outer lane of the exit guidance area are vehicles that need to exit from Exit B.

[0125] Step 4: Based on vehicle behavior and corresponding decision-making mechanisms, and combined with the collected vehicle status or distance, determine whether each type of vehicle behavior requires game theory or the definition of a game alliance. If so, proceed to Step 5.

[0126] Vehicle Behavior 1: A car enters the outer lane from Entrance A: (When a car enters the roundabout, based on the above rules, the left outer lane is designated for exiting at this intersection, meaning it does not affect vehicles entering from this intersection.) For example... Figure 7 Both entrance lanes shown have vehicles.

[0127] Decision-making mechanism 1: When 2.5≤d2-d1≤5, the two cars cooperate and play a game to enter the entrance;

[0128] When d2-d1 is not within the above range, the vehicle with smaller d1 and d2 values ​​enters the roundabout first.

[0129] Where d1 is the distance between the first vehicle (P1) in lane 1 and the edge of the circular area;

[0130] d2 is the distance between the leading vehicle (G1) in lane 2 and the edge of the circle.

[0131] Two cameras mounted on the vehicle capture stereoscopic images of the road edge. Image processing techniques are then used to calculate the distance between the vehicle and the edge.

[0132] Vehicle Behavior 2: A car in the inner lane enters the middle lane, such as... Figure 8 As shown.

[0133] Decision-making mechanism two: If there are vehicles within 5 meters of G6 in the middle lane, G6 will engage in cooperative game with these vehicles.

[0134] The methods for detecting distance include:

[0135] Radar ranging: Radar sensors emit radio waves and measure the time of flight of the reflected waves to calculate the distance between vehicles. Radar technology remains highly reliable even in adverse weather conditions.

[0136] Laser ranging: LiDAR (Light Detection and Ranging) can accurately measure the time difference between the emission and reflection of light to determine distance. This technology provides high-precision 3D environment mapping, making it ideal for use in autonomous driving systems.

[0137] Visual ranging technology: This involves using image processing algorithms to identify the size, shape, and position of vehicles, and then calculating distances based on the relative positions of the vehicles. This may involve using computer vision techniques, such as deep learning models, to identify and track vehicles.

[0138] Ultrasonic sensors: Ultrasonic sensors can be installed around a vehicle to measure the distance between the vehicle and surrounding objects (including other vehicles). This method is suitable for short-range measurements and is relatively insensitive to weather conditions.

[0139] Wireless communication technologies, such as vehicle-to-everything (V2X) technology, can transmit location and speed information between vehicles via wireless signals, thereby calculating the distance between them. This method may become more prevalent in future intelligent transportation systems.

[0140] This article uses V2X technology.

[0141] The means of determining the presence of vehicles within a specific distance is vehicle-to-everything (V2X) wireless communication. Through wireless communication between vehicles and between vehicles and infrastructure, information such as vehicle location and speed can be exchanged. This technology can provide real-time traffic information when vehicles are equipped with appropriate communication devices.

[0142] Vehicle Behavior 3: A Category G vehicle in the outer lane enters the middle lane;

[0143] Vehicle Behavior 4: R-class vehicles in the middle lane enter the inner lane.

[0144] Decision-making mechanism three applies to vehicle behavior three and vehicle behavior four, and specifically includes:

[0145] For a Class G vehicle in vehicle behavior three, if there is a vehicle of the same type behind it and the distance between the two vehicles is less than 3 meters, then the two Class G vehicles are defined as a sub-alliance. Then, for a Class R vehicle in vehicle behavior four, the sub-alliance is defined as a large alliance. This mechanism is also implemented using V2X communication technology.

[0146] Specifically, V2X communication technology can quickly obtain the driving trajectory of nearby vehicles on roundabouts and information about upcoming exits. V2X communication technology includes various communication modes such as vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), vehicle-to-pedestrian (V2P), and vehicle-to-network (V2N), which can be collectively referred to as vehicle-to-everything (V2X) communication.

[0147] When a Class G vehicle in vehicle behavior three enters buffer zone I, the vehicles within 3 meters behind the Class G vehicle, the vehicles in the middle lane parallel to the front of the Class G vehicle, and the vehicles in the outer lane within 5 meters behind the Class G vehicle and traveling in the middle lane, together form a coalition game.

[0148] Specifically, a large alliance with sub-alliances: When a Class G vehicle in the outer lane enters buffer zone I, it is determined whether the vehicle behind it is a Class G vehicle and the distance between the two vehicles is less than 3 meters. If the above conditions are met, G5 and G4 can be combined into a sub-alliance, which can then be combined with other vehicles in the middle lane to form a large alliance.

[0149] like Figure 9 The coalition game shown involves four CAVs, namely S1 = {{G5, G4}, R4, R3}. In this coalition game, G1 and G2 make the same decisions, but there may be a time delay.

[0150] Coalition Game: When a Class G vehicle from the outer lane enters Buffer Zone I, the vehicles within 3 meters behind the Class G vehicle, the vehicles in the middle lane parallel to the front of the Class G vehicle, and the vehicles within 5 meters behind the Class G vehicle from the outer lane and traveling in the middle lane, together form a coalition game. That is, S1 = {G4, G5, R3, R4}. Figure 10 As shown.

[0151] That is, when a Class G vehicle from the outer lane enters the buffer zone I, all vehicles within this range engage in a game of strategy.

[0152] If both of the above conditions occur simultaneously, calculate both and take the optimal result.

[0153] Vehicle Behavior 5: A Class P vehicle in the middle lane enters the outer lane.

[0154] Decision-making mechanism four: A coalition game is formed between the P-class vehicle and all vehicles parallel to and within 5 meters of it in the outer lane, i.e., S1 = {P2, P3, P4}. Figure 11 As shown.

[0155] That is, the P-class vehicles in the middle lane, the vehicles in the outer lane that are parallel to the front of the P-class vehicles in the middle lane, and the vehicles in the outer lane that are within 5 meters behind the P-class vehicles in the middle lane constitute the alliance game S1 in decision mechanism four.

[0156] Vehicle Behavior 6: Vehicles in the outer lane leave Exit B from Exit B.

[0157] Decision-making mechanism five: If there is a vehicle within 3 meters of lane 1, move into lane 2. Otherwise, move into lane 1. This mechanism is also implemented through V2X communication technology.

[0158] like Figure 12 As shown, when P4 approaches the turn, there is a car in lane 1, so it moves into lane 2.

[0159] Step 5: Receive the vehicle states output by all vehicle dynamics models in Step 4, combine them with driving style, and output the vehicle states after the game, that is, the vehicle states of each vehicle dynamics model under the optimal solution.

[0160] The payoff matrix in the interactive game-theoretic decision maker is as follows:

[0161]

[0162] -Driving safety cost function, - Cost function of travel efficiency - A cost function for ride comfort. 'i' represents the relevant characteristics of different driving vehicles.

[0163] All are weighting coefficients.

[0164] When the driving style is conservative It is 0.1. It is 0.4. It is 0.5;

[0165] When the driving style is aggressive It is 0.4. It is 0.5. It is 0.1;

[0166] When the driving style is stable It is 0.3. It is 0.3. It is 0.4.

[0167] Step 6: The vehicle dynamics model executes the vehicle state under the optimal solution.

[0168] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for determining lane-changing behavior of autonomous vehicles in a roundabout scenario based on hybrid game theory, characterized in that, Includes the following steps: Step 1: Under the premise of vehicle-to-everything (V2X) connectivity, capture driver behavior data and perform fuzzy inference to determine their driving style; Assume that when a vehicle enters a roundabout, the exit point is also determined and information is shared among vehicles; Driver behavior data, including: emergency braking, acceleration, turning, and vehicle speed; Emergency braking data refers to parameters of sudden emergency braking that occurs during vehicle operation, including changes in vehicle speed. Data on vehicle turning, including the steering wheel angle; Step 2: Design soft and hard rules suitable for roundabout traffic simulation; Step 3: Divide the vehicle operation area between adjacent traffic intersections of the roundabout into four categories: entrance guidance area, buffer zone I, buffer zone II, and exit guidance area. Based on soft and hard rules, establish the corresponding vehicle behavior for each type of vehicle operation area. Step 3 specifically includes the following steps: Step 3A: Divide the vehicle operation area between adjacent traffic intersections of the roundabout into four categories: entrance guidance area, buffer zone I, buffer zone II, and exit guidance area; define three types of vehicles: G, R, and P; G-class vehicles are those that ultimately leave the buffer zone through the middle lane; R-class vehicles are those that ultimately leave the buffer zone through the inner lane; and P-class vehicles are those that ultimately leave the buffer zone through the outer lane. Step 3B: Establish vehicle behavior corresponding to the operating area of ​​each type of vehicle. Vehicle behavior one and vehicle behavior two occurred in the entrance guidance area; Among them, vehicle behavior one is a car entering the outer lane from entrance A; vehicle behavior two is a car entering the middle lane from the inner lane. Vehicle behavior three and vehicle behavior four occur in buffer zone I; Vehicle behavior three: G-class vehicles in the outer lane will enter the middle lane; vehicle behavior four: R-class vehicles in the middle lane will enter the inner lane. Vehicle behavior five occurred in buffer zone II, which was a Class P vehicle in the middle lane entering the outer lane. Vehicle behavior six will occur in the exit guidance area, which is a vehicle in the outer lane leaving Exit B from Exit B. Step 4: Based on vehicle behavior and corresponding decision-making mechanisms, and combined with the collected vehicle distances, determine whether each type of vehicle behavior requires game theory or the definition of a game alliance. If so, proceed to Step 5. The decision-making mechanism in step 4 is as follows: Decision-making mechanism one, corresponding to vehicle behavior one, is as follows: when At that time, the two vehicles cooperated and negotiated to enter the entrance; when Not within the above range, and Vehicles with lower values ​​should enter the roundabout first; This refers to the distance between the leading vehicle in lane 1 and the edge of the roundabout. This refers to the distance between the leading vehicle in lane 2 and the edge of the roundabout. Decision-making mechanism two, corresponding to vehicle behavior two, is as follows: In the middle lane, if a car from the inner lane enters the middle lane and a Class G car enters the middle lane, is there any other vehicle within 5 meters? If so, the car and the other vehicles within 5 meters engage in a cooperative game. Decision-making mechanism three applies to vehicle behavior three and vehicle behavior four, specifically as follows: For a Class G vehicle in vehicle behavior three, if there is a vehicle of the same type behind it and the distance between the two vehicles is less than 3 meters, then the two Class G vehicles are defined as a sub-alliance. Then, the Class R vehicles and the sub-alliance in vehicle behavior four are defined as a large alliance. When a Class G vehicle enters Buffer Zone I in Vehicle Behavior 3, the vehicles within 3 meters behind the Class G vehicle, the vehicles in the middle lane parallel to the front of the Class G vehicle, and the vehicles in the outer lane within 5 meters behind the Class G vehicle and traveling in the middle lane, together form a coalition game. Decision-making mechanism four corresponds to vehicle behavior five, specifically as follows: The P-class vehicles in the middle lane, the vehicles in the outer lane that are parallel to the front of the P-class vehicles in the middle lane, and the vehicles in the outer lane that are within 5 meters behind the P-class vehicles in the middle lane constitute the alliance game in decision-making mechanism four. Decision-making mechanism five corresponds to vehicle behavior six, specifically: if there is a vehicle within 3 meters of lane Bo-1, then enter lane Bo-2; otherwise, enter lane Bo-1. Step 5: Receive the vehicle states output by all vehicle dynamics models in Step 4, combine them with driving style, and output the vehicle states after the game, that is, the vehicle states of each vehicle dynamics model under the optimal solution. Step 6: The vehicle dynamics model executes the vehicle state under the optimal solution.

2. The lane-changing behavior decision-making method for autonomous vehicles in a roundabout scenario based on hybrid game theory as described in claim 1, characterized in that, The specific soft and hard rules in step 2 include: Rule 1: When exiting from an adjacent intersection, travel only in the outermost lane; Rule 2: When exiting from a position one intersection away, you must first enter the middle lane, then move to the outermost lane, and finally exit. Rule 3: When exiting from a position two intersections apart, you must first enter the middle lane, then move to the outermost lane, and finally exit. Rule 4: When exiting from a position three intersections apart, you must first enter the middle lane, then the innermost lane, and then exit outwards in sequence, finally leaving the exit. Rule 5: When exiting from the entrance, you must first enter the middle lane, then the innermost lane, then enter the middle lane again, pass through two forks, then move to the outermost lane, and finally leave the exit.

3. The lane-changing behavior decision-making method for autonomous vehicles in a roundabout scenario based on hybrid game theory as described in claim 1, characterized in that, Step 1 specifically includes the following steps: Step 1A: Collect driver behavior data, including: emergency braking, acceleration, turning, and vehicle speed; Step 1B: Based on fuzzy inference, classify the collected acceleration and vehicle speed using fuzzy sets; Step 1C: Combining the fuzzy set from Step 1B, the data collected in Step 1A is transformed into driving style.

4. An autonomous driving testing system, characterized in that, include: Traffic simulator used to generate roundabout road networks; An interactive game-theoretic decision-maker, designed for simulation needs, takes the vehicle speed output from the vehicle dynamics model as input and embeds a game-theoretic decision-making algorithm based on logical rules. Considering the scenario of vehicles running in a roundabout, it divides the vehicle running area between adjacent intersections of the roundabout into four categories: entrance guidance area, buffer area I, buffer area II, and exit guidance area. It constructs a decision-making mechanism for each type of running area, determines whether the running area needs to engage in a game or define a game alliance, and if so, outputs the vehicle state after the game. An interactive game-theoretic controller is used to output control quantities for autonomous driving based on the vehicle state output by the interactive game-theoretic decision-maker. And a vehicle dynamics model, used to simulate the real response of the vehicle's mechanical system to the output of the interactive game controller, inputting throttle opening, braking pressure and front wheel deflection angle control quantities, and outputting the real vehicle state; The vehicle status includes: speed, acceleration, and driving trajectory.

Images