Non-signalized intersection traffic control method based on vehicle collaboration and gap self-adaption

Abstract

The invention discloses a non-signalized intersection traffic control method based on vehicle collaboration and gap self-adaption, and the method comprises the steps: S1, carrying out the decoupling and collaboration of the interaction between a CAV and an HDV through employing a hierarchical collaboration hybrid traffic management architecture; meanwhile, the interaction between the CAV and the pedestrian is decoupled and coordinated; s2, providing a CAV-HDV joint decision planning model fusing fuzzy logic risk perception and a game theory, converting a dynamic state of an interactive vehicle into quantitative evaluation of a human driver perception risk, and carrying out tight coupling on discrete game output and a continuous MILP-MPC motion planner; and S3, setting a cooperative control mechanism which is actively initiated by the CAV and creates a traversal gap for the pedestrian, and converting the pedestrian gap acceptance model from a passive evaluation tool to active input of CAV motion planning. According to the method, fluctuation caused by uncertainty of human behaviors is effectively stabilized, and the potential of the CAV in the mixed traffic environment is fully released.

Description

Technical Field

[0001] This invention relates to a method for traffic control at intersections, and more particularly to a method for traffic control at unsignalized intersections based on vehicle cooperation and gap adaptation. Background Technology

[0002] Unsignalized intersections are critical conflict points in urban road networks, and their operational efficiency and safety levels highly depend on autonomous negotiation and real-time decision-making among different traffic participants. With the rise of connected vehicle (CAV) technology, cooperative traffic control through vehicle-to-vehicle and vehicle-to-infrastructure communication has provided a new technological path to fundamentally improve the capacity of such intersections. In an ideal pure CAV traffic environment, centralized optimization methods based on mixed-integer linear programming (MILP) and model predictive control (MPC) have been proven to significantly reduce vehicle delays. However, real-world traffic flow will remain in a mixed state of CAVs, manually driven vehicles (HDVs), and pedestrians for a considerable period, posing a significant challenge to the practical application of cooperative control theory.

[0003] Currently, research and practice on hybrid unsignalized intersections mainly face three core bottlenecks: First, at the system-level control level, most advanced CAV cooperative control algorithms are based on the ideal assumption that all vehicles are fully controllable. When faced with HDVs whose behavior is uncertain, these algorithms become vulnerable because they cannot accurately predict their decisions, often forcing them to adopt overly conservative yielding strategies. This not only prevents CAVs from realizing their efficiency advantages, but also triggers a chain reaction due to the frequent stops of CAVs, exacerbating the disorder in the HDV traffic flow behind them and creating new sources of congestion, the so-called "reality gap" problem.

[0004] Secondly, at the vehicle-to-vehicle interaction level, although methods such as game theory are widely used to characterize right-of-way competition between CAVs and HDVs, existing models typically only output discrete decision commands such as "pass" or "yield." This results in a "decision-planning disconnect" between these commands and the underlying motion planner responsible for generating smooth, continuous trajectories. The direct consequence is that vehicles may exhibit abrupt behaviors such as sudden acceleration or deceleration, affecting not only driving comfort but also potentially causing misjudgments by surrounding HDV drivers due to the abruptness of the actions, thus creating safety hazards.

[0005] Finally, at the vehicle-pedestrian interaction level, existing research largely focuses on using machine learning or statistical models to improve the accuracy of predicting pedestrian crossing behavior. However, this "prediction-response" model is inherently passive; the vehicle-avoiding vehicle (CAV) can only passively adjust its trajectory based on the prediction results. It lacks a mechanism that allows the CAV to act as an active participant, proactively shaping and creating safe and efficient pedestrian crossing windows through its own collaborative behavior. This prevents the system from translating predictive capabilities into global efficiency and safety gains, resulting in a "passive response dilemma."

[0006] As can be seen from the above, existing technologies struggle to simultaneously achieve system-level traffic efficiency, smooth decision-making in vehicle interactions, and proactive pedestrian safety in mixed, heterogeneous unsignalized intersection environments. Specific shortcomings or deficiencies are as follows: (1) Existing collaborative control frameworks have architectural limitations and are difficult to coordinate heterogeneous traffic participants. Most existing solutions follow a single design approach, either conflating CAV, HDV, and pedestrian interactions with a unified control logic, resulting in complex models and poor adaptability; or only optimizing one type of interaction in isolation, lacking a top-level architecture that organically integrates CAV collaborative planning, HDV game theory decision-making, and pedestrian active protection. This approach leads to mutual constraints in decision-making among modules when facing concurrent conflicts of mixed traffic flows, making it impossible to achieve global optimization, and the collaborative efficiency is particularly low at low penetration rates.

[0007] (2) The existing interaction model and motion planner are loosely coupled, leading to a disconnect between decision-making and execution. Although some studies have attempted to introduce methods such as game theory to predict HDV behavior, their output is usually only an abstract policy instruction (such as "go through" or "give way"), which is not embedded in the continuous trajectory optimization model of the CAV. This creates a "semantic gap" between high-level decision-making and low-level control, forcing the vehicle to make abrupt policy transitions during the execution phase, resulting in non-smooth trajectory changes. This disjointed motion pattern not only impairs comfort but also becomes a new safety hazard because it violates the smooth expectations of human drivers.

[0008] (3) Existing pedestrian interaction mechanisms are inherently passive, and the system coordination potential of CAVs has not been activated. Current technology positions CAVs as passive responders to pedestrian behavior, with the core logic being "detection-prediction-avoidance." This model can only achieve local collision avoidance and cannot optimize traffic flow at the system level. As a controllable intelligent agent, the core advantage of CAVs lies in their ability to "create" rather than simply "find" safe gaps through proactive and coordinated maneuvers. Existing methods have failed to utilize this advantage, resulting in the system's inability to proactively resolve pedestrian-vehicle conflicts and to transform accurate behavior predictions into a positive means of improving overall traffic efficiency.

[0009] Therefore, designing an integrated cooperative traffic control method for hybrid unsignalized intersections where CAVs, HDVs, and pedestrians coexist is particularly urgent. This method aims to overcome three core shortcomings of existing technologies: disconnect between system control and reality, separation between vehicle interaction decision-making and planning, and passive response in human-vehicle interaction. Specifically, the following issues need to be addressed: (1) It solves the problem that the system-level control algorithm is vulnerable to human behavior uncertainty in mixed traffic flow and is forced to adopt conservative strategies, resulting in low efficiency. It can significantly improve traffic efficiency under both high and low CAV penetration rates.

[0010] (2) To solve the problem of disconnect between game decision-making and continuous trajectory planning in vehicle-to-vehicle interaction, by tightly coupling behavior prediction and motion planning, a safe, efficient and smooth vehicle trajectory is generated, which improves the reliability and predictability of the system.

[0011] (3) To solve the problem that CAV can only passively avoid pedestrians and cannot actively optimize in vehicle-pedestrian interaction, by transforming the pedestrian gap receiving model into an active control command, CAV can collaboratively create safe crossing opportunities, thus achieving a leap from ensuring vehicle safety to improving system efficiency. Summary of the Invention

[0012] The technical problem to be solved by the present invention is to provide a traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation, which can effectively smooth out the fluctuations caused by the uncertainty of human behavior and fully release the potential of CAV in mixed traffic environments.

[0013] To address the aforementioned technical problems, this invention provides a traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation, comprising the following steps: S1. Adopting a hierarchical and collaborative hybrid traffic management architecture, decoupling and coordinating the interaction between CAVs and HDVs, clarifying that the interaction between them is a game theory-based peer-to-peer interaction mode; simultaneously decoupling and coordinating the interaction between CAVs and pedestrians, clarifying that the interaction between them is a proactive protection-based asymmetrical protection mode; S2. Providing a CAV-HDV joint decision-making and planning model that integrates fuzzy logic risk perception and game theory, transforming the dynamic state of interacting vehicles into a quantitative assessment of the risk perceived by human drivers, and tightly coupling the discrete game output with the continuous MILP-MPC motion planner; S3. Establishing a collaborative control mechanism initiated by CAVs to create traversable gaps for pedestrians, transforming the pedestrian gap acceptance model from a passive assessment tool into an active input for CAV motion planning, and forming a "gap-adaptive vehicle group" for collaborative deceleration.

[0014] Furthermore, step S1 analyzes the management process of the mixed-traffic unsignalized intersection as follows: the spatial structure is divided into a proximity zone, a coordination control zone, and a conflict zone; real-time information sharing and collaborative planning of position, speed, and acceleration among CAVs are achieved; at the same time, CAVs obtain dynamic information of HDVs and pedestrians through onboard sensors; the movement of CAVs is uniformly coordinated and controlled by the central controller, while HDVs and pedestrians have autonomous decision-making characteristics.

[0015] Furthermore, step S1 employs the CAV collaborative motion planning method based on MILP-MPC to model the motion planning problem as a mixed integer linear programming problem, and uses model predictive control for rolling optimization.

[0016] Further, step S1 generates a globally optimized trajectory for all CAVs and performs rolling optimization using model predictive control as follows: a) State acquisition: At the beginning of each control cycle, the central controller acquires the current state of all CAVs through the communication network; b) Prediction and solution: Based on the current state, the central controller solves the mixed integer linear programming problem to obtain the optimal control sequence for all CAVs in the next time domain; c) Command issuance and execution: Only the first control quantity in the control sequence of each CAV is issued to the corresponding CAV for execution; d) Rolling advancement: At the next control cycle, steps a to c are repeated, and optimization is performed again based on the new system state.

[0017] Furthermore, step S2 employs a risk perception model based on fuzzy logic to dynamically quantify the interaction between HDV and CAV into a risk value, which is used to predict the passage intention of HDV.

[0018] Further, step S2 includes: calculating in real time the speed difference between the two parties, the distance difference to the conflict point, and the time difference to the conflict point as input values ​​for fuzzy logic; converting the input values ​​into the membership degrees of linguistic variables through a predefined membership function; and converting the output fuzzy set obtained from the reasoning into a precise driver-perceived risk value. A subset of linguistic variables; based on the calculated And HDV's driving style, predicting the threshold for aggressive drivers Threshold for normal drivers Threshold for conservative drivers The HDV traffic decision prediction process within the intersection is as follows: If If so, HDV passes through more quickly; if Then HDV maintains a constant speed; if If so, the HDV will slow down and yield.

[0019] Furthermore, in step S2, CAV maximizes overall utility based on the predicted HDV decision results in the following manner; a) Trust utility function, which refers to the utility derived from HDV's trust in CAV when the behavioral strategies of both parties reach a reciprocal state. The calculation formula is as follows: ; ; In the formula: and These represent the utility and loss of achieving and failing to achieve reciprocity in CAV and HDV behavioral strategies, respectively. and These represent the traffic strategies for vehicles, with 1 for passing and 0 for avoiding. This is a binary variable, representing whether the two parties have reached a mutually beneficial agreement.

[0020] b) Safety utility function, which refers to the utility of vehicles being able to pass through an intersection in sequence without collisions or psychological threats; it is processed using a piecewise function, and the calculation formula is as follows: ; In the formula: This refers to the acceptable safe interval between vehicle passages. This is a threshold for the increase in vehicle safety effectiveness; exceeding this value will prevent further increases in vehicle safety effectiveness. CAVs face safety losses in collisions, if The driving environment for both vehicles is safe, which will produce a positive safety effect. c) Efficiency-utility function, which refers to the utility generated by the time a vehicle takes to pass through an intersection; the calculation formula is as follows: ; ; In the formula, This indicates the passage strategy selection for CAV. This indicates the time required to pass through at maximum speed; d) CAV combined utility function; Trust utility function Safety utility function and efficiency utility function To integrate and establish a comprehensive utility function for CAV: ; In the formula, These are the weight parameters of the safety utility, efficiency utility, and trust utility functions, respectively. ; The objective function for CAV traffic decisions is to select the combination of CAV and HDV traffic strategies that maximizes the overall utility function. : .

[0021] Furthermore, step S3 divides pedestrian crossing methods into single-stage continuous crossing, two-stage crossing, and rolling gap crossing, and predicts pedestrian crossing methods as follows: first, determine the information of environmental vehicles, then calculate the maximum time for pedestrians to cross the lane, use the maximum crossing time as a basis to judge the feasibility of pedestrians accepting the gap, and finally output the crossing feasibility judgment result.

[0022] Further, step S3 determines the gap-matching vehicle group as follows: the first two vehicles in each lane are searched, meaning each matching vehicle group contains a maximum of four CAVs; the search proceeds lane by lane upstream from the conflict point, and once a CAV meeting the criteria is found in a lane, the search stops in that lane; at most one matching vehicle group exists at the same approach lane, and different vehicle groups can be planned independently in parallel; if multiple pedestrians are waiting at the same time, the matching vehicle group will prioritize serving the pedestrian with the highest probability of receiving the gap, and the gap it creates can be shared by other pedestrians; if pedestrians in two directions are located at the road edge and the central median respectively, the matching vehicle groups will be planned collaboratively to simultaneously meet the crossing needs of pedestrians in both directions.

[0023] Further, the adaptive vehicle group collaborative planning in step S3 is as follows: determine the active gap adaptation vehicle group; match pedestrian flow with crossing needs with the adaptation vehicles; generate crossing mode combinations and perform gap collaborative planning; the crossing mode combinations adopt a decision tree structure as follows: starting from the virtual root node, add branches to pedestrian flow in each direction in turn; the branching rules are as follows: prioritize enumerating from the direction of pedestrians or crowds with a higher probability of accepting the crossing gap; each direction of pedestrians corresponds to several crossing modes, and each existing leaf node is branched according to the number of its optional modes; if there are two crossing modes in a certain direction, derive two child nodes from each leaf node, each corresponding to one mode; traverse the child nodes from the root node, and finally form a complete decision tree containing four leaf nodes, covering all possible crossing mode combinations; from the root node to each leaf node is a set of candidate crossing schemes, and the crossing mode combination with the shortest active gap adaptation execution time among all crossing mode combinations is the final decision result.

[0024] Compared with the prior art, the present invention has the following beneficial effects: The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation provided by the present invention effectively mitigates the fluctuations caused by the uncertainty of human behavior by organically integrating centralized planning, distributed game decision-making and proactive service into an integrated collaborative management framework, and fully releases the potential of CAV in mixed traffic environments. Attached Figure Description

[0025] Figure 1 This is the coordinate system for the research scenario and system of this invention; Figure 2 This is a flowchart of the active gap adaptive process of the present invention; Figure 3 This is a schematic diagram of pedestrian crossing according to the present invention; Figure 4 This is a schematic diagram illustrating the pedestrian crossing method decision-making process of the present invention; Figure 5 This is a flowchart illustrating the decision-making process for the two-way pedestrian crossing method of the present invention. Detailed Implementation

[0026] The present invention will now be further described with reference to the accompanying drawings and embodiments.

[0027] This invention proposes a collaborative management framework integrating centralized planning, distributed game theory, and proactive safeguards. This framework first hierarchically decomposes the mixed traffic interactions at intersections, clarifying the different interaction paradigms and collaborative logics between CAV and HDV (game-theoretic interaction) and between CAV and pedestrians (proactive safeguards), and then constructs a unified optimization control system based on this. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation provided by this invention includes: S1. A hierarchical and collaborative hybrid traffic management architecture is proposed, decoupling and coordinating the interaction between CAVs and HDVs and between CAVs and pedestrians. This architecture is the indispensable top-level design foundation for the global optimization of this invention. It clearly distinguishes between two fundamentally different interaction modes: CAV and HDV (peer-to-peer interaction based on game theory) and CAV and pedestrians (asymmetric protection based on proactive safeguards), and designs corresponding decision-making logic. This solves the decision-making conflicts and system inefficiencies caused by existing methods that treat heterogeneous interactions as a single entity or perform isolated optimizations, ensuring the operational feasibility and traffic ethical rationality of the entire framework, and laying a systemic foundation for achieving a balance between safety and efficiency.

[0028] S2. A joint decision-making and planning model for CAV-HDV, integrating fuzzy logic risk perception and game theory, was developed. This technology is the core key to achieving smooth and safe interaction between vehicles. By constructing a fuzzy logic system, the dynamic states of interacting vehicles (speed difference, distance difference, time difference) are transformed into a quantitative assessment of the risk perceived by the human driver, making behavioral predictions more interpretable and realistic. Furthermore, this prediction result is embedded into a comprehensive utility function to guide the CAV in game-theoretic decision-making, and the discrete game output (pass / yield) is tightly coupled with a continuous MILP-MPC motion planner. This design completely solves the problem of the disconnect between "decision-planning" in traditional methods, avoids abrupt changes in vehicle motion, and ensures the executability of decisions and the stability of the system.

[0029] S3. A collaborative control mechanism was established, initiated by the CAV (Car Access Vehicle) to create crossable gaps for pedestrians. This technology is a groundbreaking innovation that significantly improves system efficiency and is indispensable. It fundamentally changes the role of the CAV in vehicle-pedestrian interaction, transforming it from a passive "predictor-responder" to an active "traffic flow organizer." By transforming the pedestrian gap acceptance model from a passive assessment tool into an active input for CAV motion planning, and by forming a "gap-adaptive vehicle group" for coordinated deceleration, this invention can dynamically "create" safe and efficient pedestrian crossing windows. This mechanism successfully transforms the predictive insight into pedestrian behavior into global traffic efficiency benefits, effectively solving the long-standing "passive response dilemma," and significantly reducing the total delay caused by mutual waiting between pedestrians and vehicles while ensuring absolute pedestrian safety.

[0030] The specific implementation process of this invention mainly includes the following core steps: S11: Analysis of the management process at a mixed-traffic, unsignaled intersection, as detailed below: This invention is applicable to unsignalized intersections where CAVs, HDVs, and pedestrians share the same space, as shown in the following description. Figure 1 As shown, the system is mainly divided into three functional areas: the proximity zone, the coordination and control zone, and the conflict zone. In this system, CAVs can share real-time information on their position, speed, acceleration, and other statuses, and coordinate planning. Simultaneously, CAVs can also acquire dynamic information about HDVs and pedestrians through onboard sensors. The movement of CAVs is uniformly coordinated and controlled by the central controller, while HDVs and pedestrians have autonomous decision-making characteristics, and their behavior is not directly controlled by the system, thus forming a typical mixed-traffic, heterogeneous traffic environment.

[0031] In this mixed traffic scenario, the decision-making behaviors and traffic patterns of the three types of traffic participants differ significantly: As controlled objects, CAVs exhibit highly predictable and coordinated behavior, enabling efficient passage through centralized planning.

[0032] HDVs are controlled by drivers, and their behavior is influenced by individual driving style and real-time traffic conditions. They exhibit strong autonomy and uncertainty, and are prone to behaviors such as cutting in or hesitating at conflict points.

[0033] Pedestrians prioritize safe crossing, and their decisions are influenced by factors such as crossable gaps and waiting times. They typically decide whether to cross based on gap acceptance models. Pedestrian crossings often employ single-cross, double-cross, or rolling gap crossing patterns, causing intermittent interference with traffic flow.

[0034] Step S12: The CAV cooperative motion planning method based on MILP-MPC, the specific process is as follows: This step aims to generate globally optimized, safe, and smooth trajectories for all CAVs. Its core is to model the motion planning problem as a mixed-integer linear programming problem and employ model predictive control for rolling optimization to handle dynamic changes in the system. Specifically: (1) Vehicle kinematic constraints; The discretized vehicle kinematics equations are as follows: (1) (2) (3) (4) (5) (6) In the formula: Indicates the number of discrete-time steps; ; lane CAVs set on ; Represents the set of all lanes; lane vehicles on exist The coordinates of time; This is the simulation time step; lane vehicles on exist The speed of time; Indicates the vehicle's maximum speed; lane vehicles on exist Acceleration at any moment; and These represent the maximum and minimum accelerations, respectively. lane vehicles on exist The rate of change of acceleration at any given moment; and These represent the minimum and maximum rates of change of acceleration, respectively.

[0035] (2) The safety constraint equations are as follows: Constraint (7) guarantees the vehicle With the car in front Maintain a safe distance; constraint (8) ensures lane vehicles on Conflict lane Get on the vehicle Maintain a minimum safe distance between vehicles to avoid collisions and ensure that one vehicle has safely passed before the other reaches the point of conflict, as detailed below: (7) (8) In the formula: Indicates the length of the vehicle; Indicates the minimum safe distance between vehicles; Reaction time; moment , lane vehicles on All vehicles ahead must assemble, including CAVs and HDVs; and the lane must be cleared. The set of lanes with conflict is , ; For lane vehicles on The coordinates; For lane Conflicting lane The point of conflict to the lane superior Horizontal distance of points; For lane Conflicting lane The point of conflict to the lane superior The horizontal distance between points.

[0036] (3) Linearization process: To achieve efficient solution, this invention introduces slack variables. and The above non-convex constraints are transformed into the following set of linear constraints, and the problem is transformed into MILP: Constraints (7) and (8) are rewritten in the following form: (9) (10) (11) The modified constraint (17) is as follows: (12) (13) (14) (15) (16) (17) (4) Optimize the objective function: The goal of the central controller is to maximize traffic efficiency while ensuring comfort and planning feasibility. The objective function is designed as follows: (18) (19) (20) ;(twenty one) ;(twenty two) ;(twenty three) In the formula: , and It is a non-negative auxiliary variable; It is a very large value; It is a very large constant that can guarantee The minimum value is obtained to avoid excessively slack safety constraints.

[0037] (5) Model Predictive Control (MPC) Rolling Optimization; a) Status acquisition: at the beginning of each control cycle The central controller obtains the current status (position, velocity, acceleration) of all CAVs through the communication network.

[0038] b) Prediction and Solution: Based on the current state, the controller solves the above MILP problem to obtain a future time domain value. The optimal control sequence (acceleration sequence) for all CAVs.

[0039] c) Command Issuance and Execution: Only the first control variable in each CAV control sequence (i.e., ...) is executed. arrive The acceleration is sent to the corresponding CAV for execution.

[0040] d) Rolling forward: to the next control cycle Repeat steps a to c, and re-optimize based on the new system state.

[0041] This MPC framework enables the system to continuously adapt to changes in traffic flow and is robust to prediction errors and external disturbances.

[0042] S2 is a CAV-HDV interactive decision-making model based on fuzzy logic and game theory; This step is central to addressing mixed traffic uncertainty. It enables CAVs to proactively predict and respond to the autonomous decisions of HDVs, thereby incorporating the uncertainty of human behavior into the collaborative planning framework.

[0043] S21, HDV Behavior Prediction: Risk Perception Model Based on Fuzzy Logic; This module uses a fuzzy logic system to dynamically quantify the interaction between HDV and CAV into a risk value, which is used to predict the passage intention of HDV.

[0044] a) Input variables for calculation; The system calculates three key interaction metrics in real time as inputs for fuzzy logic: The formula for calculating the absolute value of the speed difference is as follows: ;(twenty four) The formula for calculating the distance difference to the conflict point is as follows: (25) In the formula, and This represents the distance between the vehicle's current location and the point of conflict.

[0045] absolute value of time difference The calculation is relatively complex because the exact time it takes for the vehicle to reach the conflict point cannot be accurately obtained. The arrival time at the conflict point can be categorized into two scenarios: acceleration and deceleration. The arrival time during acceleration can also be divided into two cases: one where the vehicle has not reached its maximum speed when reaching the conflict point. At this point, it is assumed that the vehicle will maintain uniform acceleration during this motion; secondly, the vehicle will first accelerate for a period of time until it reaches its maximum speed. Then maintain maximum speed through the conflict point. At this time, the lane ( Vehicles on ( Time of arrival at the point of conflict ( The formula is as follows: (26) (27) In the formula, It is the longest distance a vehicle can travel during acceleration.

[0046] There are two scenarios when a vehicle decelerates: either it continues to decelerate until it reaches the conflict point, or it stops before reaching the conflict point. The formula for the time to reach the conflict point is as follows: (28) (29) In formula (28), This indicates that the vehicle will stop before reaching the conflict point. Under the current deceleration and avoidance strategy, the time it takes for the vehicle to reach the conflict point is considered infinite, hence a very large constant value is used. This indicates the arrival time of the vehicle.

[0047] Calculate the time taken for CAVs and HDVs to reach the conflict point using formulas (26)-(29). and The formula for calculating the absolute value of the time difference between the arrival of both sides at the point of conflict is as follows: (30) b) Fuzzy reasoning process; Blur: Divide the three , , Precise input values ​​are transformed into membership degrees of linguistic variables using predefined membership functions (such as triangular and trapezoidal functions), as shown in Table 1: Table 1. Fuzzy logic rule settings:

[0048] Defuzzification: Transforming the fuzzy output set obtained from inference into a precise driver-perceived risk value. The linguistic variable subsets are: maximum (PH), large (H), medium (ME), small (SM), and minimum (NS). A higher value indicates a higher level of danger perceived by the HDV driver.

[0049] c) HDVs decision prediction; According to the calculation Based on the HDV's driving style, predict its decision-making: Aggressive drivers (threshold) ); Normal type driver (threshold) ); Conservative driver (threshold) ); The HDV traffic decision prediction process within the intersection is as follows: 1) If 1) HDV passes through faster; 2) If 3) If HDV maintains a constant speed; If the HDV slows down to give way, then after the decision is made, the HDVs update their velocity for the next time step according to formula (48). (31) S22, CAV game decision-making: maximizing overall utility; CAV determines its optimal strategy by solving an optimization problem based on the predicted HDV decision results.

[0050] a) Trust utility function; Trust utility refers to the utility derived from HDV's trust in CAV when both parties' behavioral strategies reach a mutually beneficial state (one party proceeds while the other avoids). Otherwise, if the behavioral strategies of both parties do not reach a mutually beneficial state, a loss occurs. The trust utility function is shown in formula (32): (32) (33) In the formula: and These represent the utility and loss of achieving and failing to achieve reciprocity in CAV and HDV behavioral strategies, respectively. and These represent the traffic strategies for vehicles, with 1 for passing and 0 for avoiding. This is a binary variable, representing whether the two parties have reached a mutually beneficial agreement.

[0051] b) Safety utility function; Safety utility refers to the benefit of vehicles being able to pass through an intersection in sequence without collisions or psychological threats. Because safety utility exhibits diminishing marginal utility, a piecewise function is used to process it. The calculation formula is as follows: (34) In the formula: This refers to the acceptable safe interval between vehicle passages. This is a threshold for the increase in vehicle safety effectiveness; exceeding this value will prevent further increases in vehicle safety effectiveness. CAVs face safety losses in collisions, if If the driving environment for both vehicles is safe, then a positive safety effect will be generated.

[0052] c) Efficiency-utility function; Efficiency utility refers to the utility generated by the time a vehicle takes to pass through an intersection. If a CAV chooses to proceed, it gains efficiency utility related to its own passage time; otherwise, it incurs an efficiency loss due to the waiting time incurred to yield to the oncoming vehicle. This loss is related to the oncoming vehicle's passage time through the conflict point. The efficiency utility function formula is as follows: (35) (36) In the formula, This indicates the passage strategy selection for CAV. This indicates the time required to pass through at maximum speed.

[0053] d) CAV combined utility function; Trust utility function Safety utility function and efficiency utility function To integrate and establish a comprehensive utility function for CAV: (37) In the formula, These are the weight parameters of the safety utility, efficiency utility, and trust utility functions, respectively. .

[0054] The objective function for CAV's traffic decision is as follows: select the combination of CAV and HDV traffic strategies that maximizes the overall utility function. : (38) S3 Active gap adaptation model for pedestrian crossing; This step is the core innovation of this invention in protecting vulnerable road users. It transforms the role of CAVs from a passive reactant to an active traffic flow organizer, "creating" safe crossing opportunities for pedestrians through coordinated maneuvering. The overall workflow of this model is as follows: Figure 2 As shown, its core components include: pedestrian crossing behavior prediction, gap matching vehicle group determination, and collaborative motion planning.

[0055] S31. Pedestrian crossing pattern prediction; This step addresses pedestrian crossing behavior at unsignalized intersections, categorizing crossing methods into three basic types: single-stage continuous crossing, two-stage crossing (i.e., completing the second stage crossing after waiting in the central stopping area), and rolling gap crossing (i.e., making decisions lane-by-lane and utilizing available gaps). For medium-to-high traffic flow scenarios more common in actual traffic management, this invention focuses on solving the collaborative control problem when pedestrians cross using two-stage crossing and rolling gap crossing methods.

[0056] To ensure the effectiveness of gap adaptation, this invention explicitly incorporates the calculation of the time required for a pedestrian to cross a single lane into the control model, instead of making microscopic predictions of the pedestrian trajectory, and simplifies the pedestrian crossing motion into a uniform or uniformly accelerated linear motion model.

[0057] pedestrian The moment begins, two lanes are crossed consecutively. and The formula for calculating the second crossing time is as follows: (39) pedestrian The moment begins, one lane at a time. and lane The formula for calculating the crossing time of the rolling gap is as follows: (40) (41) (41) (42) In the formula, pedestrian crossing pedestrians on ; For the width of one lane, , and This is for time travel, because the process of crossing the street in the second time travel method is continuous, therefore And the rolling gap crosses the street .

[0058] A probabilistic model for pedestrians accepting and rejecting a crossing gap in a multi-lane road system was established based on logistic regression. The probabilities of accepting and rejecting a crossing gap are expressed as follows: and The probability that a pedestrian will accept a gap is as follows: (43) (44) In the formula, , , It is the base of the natural logarithm, when near hour, Approaching 0, when near hour, Approaching 1. A value close to 0 indicates that the probability of an event occurring is very low, while a value close to 1 indicates that the probability of an event occurring is very high. (seconds) indicates the time required for a vehicle to travel at its current speed to the point of conflict at the pedestrian crossing. (meters) indicates the distance from the edge of the road (or the center line of the road) to the far side of the road. (seconds) indicates the pedestrian waiting time. (vehicles / hour) indicates the number of vehicles passing through the road section per hour. This indicates the number of pedestrians waiting to cross the street on the side or in the middle of the road. This indicates whether pedestrians are waiting to cross the street at the edge or in the center. Pedestrians often feel it is more dangerous to wait in the middle of the road and will choose to cross through the smaller gap. It is a binary discrete variable, with 1 and 0 representing waiting at the edge and in the center of the road, respectively.

[0059] In the above formula, The time required for pedestrians to cross the road was not taken into account, but in reality, pedestrians need time to cross the lane. A simple correction has been made, which has four advantages: 1) The feasibility of pedestrian crossing gaps and the judgment of crossing methods are more accurate; 2) The correction of crossing gaps takes into account the influence of lanes; 3) The feasible solution space for CAVs active gap adaptation is reduced, and the solution efficiency is improved; 4) Pedestrian safety is more fully guaranteed.

[0060] (45) In the formula, and These represent vehicles in lane 1 and lane 2, respectively. The time required to reach the pedestrian crossing conflict point at the current speed; and These represent vehicles in lane 1 and lane 2, respectively. The corrected crossing gap that can be provided; and These represent the time it takes for pedestrians to cross lanes 1 and 2, respectively. and These represent the minimum acceptable clearances on lane 1 and lane 2, respectively. and These represent the corrected minimum acceptable clearance on lane 1 and lane 2, respectively.

[0061] by A value greater than 0.5 indicates that the pedestrian acceptance gap is the standard. The above model has an accuracy rate of 96.77%, and the accuracy rate is further improved after the upgrade. This model is used to determine the feasibility of pedestrians crossing a single lane. In a mixed-traffic, unsignaled intersection environment, the pedestrian crossing mode decision-making process is as follows: Figure 4 As shown.

[0062] Based on the above process, by repeatedly using the model, it can be determined whether a pedestrian should use a two-stage crossing or a rolling gap crossing. When a pedestrian reaches the central divider, he / she will face lanes 3 and 4 in the opposite direction, and the decision-making process is the same as that for lanes 1 and 2. In summary, the specific process for determining the feasibility of accepting the gap crossing and the appropriate crossing method is as follows: 1. First, determine the information about the vehicles in the environment (speed, position, acceleration, and vehicle type of the first two vehicles in each lane); 2. Then calculate the maximum time for pedestrians to cross the lane (the feasibility of pedestrians accepting the gap is judged based on the maximum crossing time, which can not only fully ensure the safety of pedestrians, but also prevent pedestrians from colliding with vehicles when they are forced to cross the street after waiting too long). 3. Based on the flowchart above, determine which method the pedestrian chooses to cross the street; 4. Output the feasibility assessment result of the crossing (first crossing, second crossing, rolling gap crossing, or waiting for feasibility).

[0063] S32. Dynamic construction and task allocation of gap-adaptive vehicle groups; After determining the pedestrian crossing method, the system will construct a "gap-adaptive vehicle group," which is a set of CAVs that need to coordinate their movement to provide a safe crossing gap for pedestrians. For example... Figure 1 As shown, the vehicle groups A and B are a combination of CAVs that cooperate across lanes.

[0064] To determine the appropriate gap for the vehicle group, this invention establishes the following rules: (1) Each lane can search for a maximum of the first 2 vehicles, that is, each matching vehicle group can contain a maximum of 4 CAVs; (2) Search upstream from the point of conflict, lane by lane. Once a CAV that meets the criteria is found in a lane, stop searching in that lane.

[0065] pedestrians For example, its compatible vehicle group A consists of CAVs: and Composition. If the first vehicle in a lane is an HDV (e.g., lane 4), If the upstream adjacent vehicle is determined, then the judgment continues (e.g., If it is a CAV, it is included in the vehicle group; otherwise, the lane does not meet the matching conditions.

[0066] During system operation, only one compatible vehicle group can exist at a time on the same approach lane, while different vehicle groups can be planned independently and in parallel. If multiple pedestrians are waiting simultaneously, the compatible vehicle group will prioritize serving the pedestrian with the most urgent intention to cross the street (highest probability of accepting the gap), and the gap it creates can be shared by other pedestrians. Furthermore, if pedestrians from two directions are located at the road edge and the central median respectively (e.g.,...), the vehicle group will prioritize serving the pedestrian with the most urgent intention to cross the street (highest probability of accepting the gap), and the gap created by the vehicle group can be shared by other pedestrians. and ), then it is suitable for vehicle groups ( and Collaborative planning is needed to simultaneously meet the crossing needs of pedestrians in both directions.

[0067] S33. An active gap adaptive method for one-way pedestrian crossing needs; The essence of active clearance adaptation lies in dynamically adjusting the movement of CAVs (Car Access Control Systems) by sensing the pedestrian's movement status in real time, while ensuring safety, thereby determining the optimal clearance adaptation scheme. However, not all CAVs can effectively perform active clearance adaptation. For example, when the vehicle speed is too high and the necessary movement adjustment cannot be completed before reaching the conflict point, the adaptation behavior is not feasible. This means that clearance adaptation can only be performed when the CAV's movement state meets specific constraints, that is, its feasible state set must be defined first.

[0068] pedestrians Taking crossing lane 1 and lane 2 as an example, after the two CAVs have completed their active clearance matching, the pedestrian begins to cross. At this time, the pedestrian's motion state is as follows: The feasible state set of CAVs is determined by CAVs detection. and , recorded as .when and Given, then and These are two sets. When pedestrians use two-stage crossing and rolling gap crossing, the vehicle status needs to meet certain constraints to ensure pedestrian safety.

[0069] A. For a two-stage street crossing, there is no stopping or waiting process during the crossing, so the pedestrian's acceleration is the same throughout the entire process. The feasible state constraints for the two CAVs to actively create a crossing gap for the pedestrian are as follows: (46) (47) (48) (49) (50) (51) (52) (53) (54) (55) B. For rolling gap crossings, there will be a stopping and waiting process during the crossing, so the initial velocity and acceleration of the pedestrian are different in the two crossing processes. The feasible state constraints for the two CAVs to actively create crossing gaps for the pedestrian are as follows: (56) (57) (58) (59) (60) (61) (62) (63) (64) Given a fixed pedestrian position and velocity, the above formula specifies the range of values ​​for the adaptive vehicle acceleration and velocity, i.e., the state set at the moment the pedestrian begins crossing. The state set indicates the conditions under which CAVs can actively create gaps to meet pedestrian crossing needs. Therefore, the decision variables for the active gap adaptation model are the motion states of the adapting CAVs on the two lanes. Since the vehicle position change can be calculated given the speed, and the speed can be obtained from the acceleration, the model decision variables are... and Based on the motion planning model of intelligent connected vehicles, the objective function of the active clearance adaptation model of CAVs is as follows: (65) (66) The objective function represents minimizing the total time from when the vehicle begins active gap adaptation to when the pedestrian completes crossing, while ensuring both comfort and efficiency; that is, minimizing the vehicle loss time caused by pedestrian interference.

[0070] S34. An active gap adaptive method for bidirectional pedestrian crossing needs; The aforementioned one-way clearance adaptation model provides a solution for handling pedestrian crossings in one direction. However, in actual intersections, situations often arise where pedestrian flows in both directions need to cross the same traffic flow, such as... Figure 5 As shown in the diagram, pedestrians in each direction can choose between a rolling gap or a secondary crossing method, resulting in a variety of possible crossing combinations. Different combinations correspond to different pedestrian crossing times, which in turn affect the constraints of CAVs and the execution time of gap planning.

[0071] Therefore, coordinating the crossing needs of pedestrians in both directions and determining the optimal adaptation scheme becomes the core task of the collaborative planning model. This invention proposes the following three-step collaborative planning process: (1) Determine the vehicle group that adapts to the active clearance; (2) Match pedestrian flow with crossing needs with suitable vehicles; (3) Generate combinations of crossing methods and perform gap collaborative planning.

[0072] To systematically enumerate possible combinations of bidirectional pedestrian crossing methods, this invention employs a decision tree structure. This decision tree constructs a complete set of candidate solutions by traversing all feasible combinations of pedestrian crossing methods, as shown in the following structure: Figure 5 As shown.

[0073] The decision tree construction process is as follows: Starting from the virtual root node, branches are added sequentially for pedestrian flow in each direction. The branching rules are as follows: 1. Prioritize enumerating from the direction of pedestrians or crowds with a higher probability of accepting passage through gaps, for example, starting with pedestrians. This is the first branch direction; 2. Pedestrians in each direction have several crossing methods (this invention considers two methods: rolling gap and secondary crossing), and each existing leaf node is branched and expanded according to the number of its selectable methods; 3. If there are two ways to traverse a certain direction, then derive two child nodes from each leaf node, each corresponding to one of the ways.

[0074] Starting from the root node The first branch is " (Scrolling) and " (Secondary) two child nodes; then Building upon this foundation, we can further expand upon it, starting from... Each of the two child nodes derives " (Scrolling) and " (Secondary) nodes ultimately form a complete decision tree containing four types of leaf nodes, covering all possible combinations of traversal methods.

[0075] From the root node to each leaf node, there is a set of candidate crossing options. The decision tree for combining crossing methods constructs all possible crossing options. For example, Figure 5 The number of nodes from the root to the leaf node in the decision tree for crossing modes is 4, therefore there are 4 crossing schemes, as shown in Figure 5. Different crossing schemes have different impacts on CAVs. This invention aims to minimize the interference time with CAVs by performing collaborative planning and decision-making for crossing mode combinations; adapting to vehicles. and The HDV at the rear updates its state according to its corresponding motion model. The decision-making process for the optimal combination of pedestrian crossing methods in both directions is as follows:

[0076] In summary, among all combinations of traversal methods, the combination with the shortest active gap adaptation execution time is the final decision result. .

[0077] The technical problems that this invention can solve include the following: (1) A centralized CAV cooperative motion planner based on MILP-MPC was constructed, which solved the problem of disconnect between system-level control and real traffic. The planner establishes a mixed-integer linear programming model that includes vehicle dynamics, safety distance and conflict avoidance constraints, and uses model predictive control (MPC) for rolling time-domain optimization to generate a globally optimized smooth trajectory for the CAV fleet. Its innovation lies in the introduction of slack variables and linearization techniques, which ensures the feasibility and efficiency of solving the problem when there are new vehicle interferences and nonlinear conflict constraints, thereby overcoming the shortcomings of existing centralized algorithms that are prone to failure in mixed traffic flow and laying the foundation for efficient traffic flow.

[0078] (2) A CAV-HDV interactive decision-making model integrating fuzzy logic and game theory was designed to solve the problem of disconnect between the decision-making and planning stages. This model first uses a fuzzy logic system to transform the relative speed, relative distance, and expected conflict time of the interacting vehicles into a quantitative assessment of the risk perceived by the human driver, achieving accurate prediction of the HDV's passage intention. Subsequently, based on this prediction, a comprehensive utility function considering trust, safety, and efficiency is established to guide the CAV in game-theoretic decision-making, and the final strategy selection is seamlessly embedded into the aforementioned MILP-MPC motion planner. This tightly coupled mechanism ensures that discrete interactive decisions can be directly and smoothly transformed into continuous and safe vehicle trajectories, eliminating the safety hazards caused by decision jumps.

[0079] (3) A proactive gap adaptation mechanism for pedestrian crossings is proposed, solving the dilemma of passive response in vehicle-pedestrian interaction. The core of this mechanism is to transform CAVs from passive yielders to proactive traffic flow coordinators. When a pedestrian crossing demand is detected, the system dynamically assembles a "gap-adaptive CAV fleet" and uses an improved pedestrian gap acceptance probability model to evaluate the feasibility of crossing. Furthermore, with the goal of generating traffic gaps that meet the safe crossing time for pedestrians, the cooperative motion constraints of the CAV fleet are established and integrated into the central planner as an optimization objective. This enables CAVs to proactively create safe and efficient crossing windows for pedestrians through mechanisms such as cooperative deceleration, thereby transforming accurate behavior prediction into proactive control actions that improve the overall efficiency of the system.

[0080] The cooperative passage and active gap adaptation management framework proposed in this invention has produced significant beneficial effects in solving the problem of coordinating efficiency and safety optimization at mixed unsignalized intersections. Its most core beneficial effect is that, through the cooperative control of CAVs, it systematically breaks the deadlock of the traditional control strategy where efficiency and safety are at odds, achieving a significant reduction in both vehicle and pedestrian delays and significantly improving traffic flow stability.

[0081] Specifically, through large-scale simulation experiments, this technical solution has produced the following direct and quantifiable beneficial effects: In terms of traffic efficiency, this invention achieves a significant reduction in vehicle and pedestrian delays. Compared to traditional conservative yielding strategies, this invention effectively reduces average delays at various CAV penetration rates. The effect is most significant at high penetration (90%) scenarios: average vehicle delays decrease from 26.3 seconds in the baseline model to 16.5 seconds, a reduction of 37.3%; average pedestrian delays decrease from 18.9 seconds to 11.8 seconds, a reduction of 37.6%. Even at low penetration (10%), delay reductions of 8.4% and 8.7% are achieved, respectively. This demonstrates that this method delivers substantial efficiency gains throughout the entire CAV adoption process.

[0082] In terms of congestion relief, this invention effectively suppresses the growth of intersection queue lengths. Through collaborative optimization and active gap creation, it smooths traffic flow and reduces unnecessary stop-and-go starts. Experimental results show that at a 90% CAV penetration rate, the average vehicle queue length decreased from 4.9 vehicles in the baseline model to 2.6 vehicles, a reduction of 46.9%; the average pedestrian queue length decreased from 2.6 people to 1.5 people, a reduction of 42.3%. This significant reduction in queue length directly lowers the risk of intersection spillover and improves the overall capacity of the road network.

[0083] In terms of system stability, the predictability and robustness of traffic flow are significantly enhanced. Through accurate HDV behavior prediction and proactive traffic flow organization, this invention effectively mitigates fluctuations caused by the uncertainty of human behavior. Data shows that with the increase in CAV penetration, the standard deviation of vehicle and pedestrian delays decreases more significantly after adopting this invention compared to the baseline model, with an average stability improvement of approximately 25%-30%. This means that traffic operations become more reliable, enabling accurate travel time prediction and management.

[0084] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications and improvements without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be defined by the claims.

Claims

1. A method for traffic control at unsignalized intersections based on vehicle cooperation and gap adaptation, characterized in that, Includes the following steps: S1. Adopt a layered and collaborative hybrid traffic management architecture to decouple and coordinate the interaction between CAV and HDV, clarifying that the interaction between the two is a peer-to-peer interaction mode based on game theory; at the same time, decouple and coordinate the interaction between CAV and pedestrians, clarifying that the interaction between the two is a non-peer-to-peer protection interaction mode based on proactive protection. S2. A CAV-HDV joint decision-making and planning model that integrates fuzzy logic risk perception and game theory is provided. The dynamic state of the interactive vehicle is transformed into a quantitative assessment of the risk perceived by the human driver, and the discrete game output is tightly coupled with the continuous MILP-MPC motion planner. S3. Establish a collaborative control mechanism initiated by CAV to create crossable gaps for pedestrians, transform the pedestrian gap acceptance model from a passive evaluation tool into an active input for CAV motion planning, and form a gap-adaptive vehicle group for collaborative deceleration.

2. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 1, characterized in that, Step S1 analyzes the management process of the mixed-traffic unsignalized intersection as follows: the spatial structure is divided into a proximity zone, a coordination control zone, and a conflict zone; real-time information sharing and collaborative planning of position, speed, and acceleration among CAVs are achieved; at the same time, CAVs obtain dynamic information of HDVs and pedestrians through onboard sensors; the movement of CAVs is uniformly coordinated and controlled by the central controller, while HDVs and pedestrians have autonomous decision-making characteristics.

3. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 1, characterized in that, Step S1 employs the CAV collaborative motion planning method based on MILP-MPC to model the motion planning problem as a mixed integer linear programming problem, and uses model predictive control for rolling optimization.

4. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 3, characterized in that, Step S1 generates a globally optimized trajectory for all CAVs and performs the following rolling optimization using model predictive control: a) Status acquisition: At the beginning of each control cycle, the central controller acquires the current status of all CAVs through the communication network; b) Prediction and solution: Based on the current state, the central controller solves the mixed-integer linear programming problem to obtain the optimal control sequence for all CAVs in the next time domain; c) Command issuance and execution: Only the first control variable in each CAV control sequence is issued to the corresponding CAV for execution; d) Rolling progress: In the next control cycle, repeat steps a to c, and re-optimize according to the new system state.

5. The method for traffic control at unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 1, characterized in that, Step S2 employs a risk perception model based on fuzzy logic to dynamically quantify the interaction between HDV and CAV into a risk value, which is used to predict the passage intention of HDV.

6. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 5, characterized in that, Step S2 includes: The three key interactive indicators—the speed difference between the two sides, the distance difference to the conflict point, and the time difference to the conflict point—are calculated in real time and used as input values ​​for the fuzzy logic. The input values ​​are transformed into the membership degrees of linguistic variables using a predefined membership function; The fuzzy output set obtained from the inference is transformed into a precise driver-perceived risk value. A subset of language variables; According to the calculation And HDV's driving style, predicting the threshold for aggressive drivers Threshold for normal drivers Threshold for conservative drivers ); The HDV traffic decision prediction process within the intersection is as follows: If If so, HDV passes through more quickly; if Then HDV maintains a constant speed; if If so, the HDV will slow down and yield.

7. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 1, characterized in that, In step S2, CAV maximizes overall utility based on the predicted HDV decision results in the following manner; a) Trust utility function, which refers to the utility derived from HDV's trust in CAV when the behavioral strategies of both parties reach a reciprocal state. The calculation formula is as follows: ； ； In the formula: and These represent the utility and loss of achieving and failing to achieve reciprocity in CAV and HDV behavioral strategies, respectively. and These represent the traffic strategies for vehicles, with 1 for passing and 0 for avoiding. This is a binary variable, representing whether the two parties have reached a mutually beneficial agreement; b) Safety utility function, which refers to the utility of vehicles being able to pass through an intersection in sequence without collisions or psychological threats; it is processed using a piecewise function, and the calculation formula is as follows: ； In the formula: This refers to the acceptable safe interval between vehicle passages. This is a threshold for the increase in vehicle safety effectiveness; exceeding this value will prevent further increases in vehicle safety effectiveness. CAVs face safety losses in collisions, if The driving environment for both vehicles is safe, which will produce a positive safety effect. c) Efficiency-utility function, which refers to the utility generated by the time a vehicle takes to pass through an intersection; the calculation formula is as follows: ； ； In the formula, This indicates the passage strategy selection for CAV. This indicates the time required to pass through at maximum speed; d) CAV combined utility function: Trust utility function Safety utility function and efficiency utility function To integrate and establish a comprehensive utility function for CAV: ； In the formula, These are the weight parameters of the safety utility, efficiency utility, and trust utility functions, respectively. ; The objective function for CAV traffic decisions is to select the combination of CAV and HDV traffic strategies that maximizes the overall utility function. : .

8. The method for traffic control at unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 1, characterized in that, Step S3 divides pedestrian crossing methods into single-stage continuous crossing, two-stage crossing, and rolling gap crossing, and predicts pedestrian crossing methods as follows: first, determine the information of environmental vehicles, then calculate the maximum time for pedestrians to cross the lane, use the maximum crossing time as a basis to judge the feasibility of pedestrians accepting the gap, and finally output the crossing feasibility judgment result.

9. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 1, characterized in that, Step S3 determines the gap-fitting vehicle group as follows: Each lane can search for a maximum of the first 2 vehicles, meaning each matching vehicle group can contain a maximum of 4 CAVs; The search proceeds upstream from the point of conflict, lane by lane. Once a CAV that meets the criteria is found in a lane, the search stops in that lane. At most one vehicle group can exist at the same entrance lane at the same time. Different vehicle groups can be planned independently in parallel. If multiple pedestrians are waiting at the same time, the vehicle group will prioritize serving the pedestrian with the highest probability of receiving the gap, and the gap it creates can be shared by other pedestrians. If pedestrians in two directions are located at the road edge and the central median respectively, then vehicle group collaborative planning should be adapted to simultaneously meet the crossing needs of pedestrians in both directions.

10. The traffic control method for unsignalized intersections based on vehicle cooperation and gap adaptation as described in claim 9, characterized in that, The adaptive vehicle group collaborative planning in step S3 is as follows: determine the active gap adaptive vehicle group; match pedestrian flow with crossing needs with adaptive vehicles; generate crossing mode combinations and perform gap collaborative planning; The combination of traversal methods is configured using a decision tree structure as follows: Starting from the virtual root node, branches are added to the pedestrian flow in each direction in turn; the branching rules are as follows: priority is given to enumerating the direction of pedestrians or crowds with a higher probability of accepting the crossing gap; each direction of pedestrians corresponds to several crossing methods, and each existing leaf node is branched and expanded according to the number of its optional methods; if there are two crossing methods in a certain direction, two child nodes are derived from each leaf node, each corresponding to one method. Starting from the root node, traverse the child nodes to eventually form a complete decision tree containing four types of leaf nodes, covering all possible combinations of traversal methods; From the root node to each leaf node, there is a set of candidate crossing schemes. Among all the crossing scheme combinations, the crossing scheme combination with the shortest active gap adaptation execution time is the final decision result.

Images