Multi-vehicle cooperative driving double-layer decision optimization method in intelligent connected hybrid traffic environment

Abstract

The application discloses a kind of multi-vehicle cooperative driving double-layer decision optimization methods under intelligent network connection hybrid traffic environment, comprising: adjusting lane type distribution, selecting target lane for vehicle;Quantify the driving intention of vehicle, behavior search and target area allocation are carried out, and the driving decision of vehicle is obtained;Based on driving decision, make each intelligent network connection automobile from current position smooth to target area.This application can realize adaptive modeling and real-time optimization to multi-vehicle interactive behavior, improve the traffic efficiency under complex road conditions and the stability of overall traffic.

Description

Technical Field

[0001] This invention relates to the field of vehicle traffic control, specifically to a two-level decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment. Background Technology

[0002] With the deep integration of intelligent connected vehicle (V2X) technology and autonomous driving systems, transportation systems are evolving towards greater coordination and real-time optimization. Especially in urban arterial roads and highways where multi-vehicle interactions are dense and decision-making is complex, various types of vehicles, such as intelligent connected vehicles (ICVs) and manually driven vehicles (HDVs), need to frequently interact and coordinate. How to achieve efficient lane allocation and vehicle scheduling has become a key factor affecting overall traffic efficiency and safety levels.

[0003] However, existing traffic control methods largely rely on centralized optimization or static signal control, which struggles to respond promptly to sudden road conditions and dynamic changes such as lane resource shortages, easily leading to traffic bottlenecks, delays, and even safety risks. Furthermore, traditional models are generally based on the assumption of homogeneous traffic flow, lacking a detailed characterization of the characteristics and behavioral differences of different vehicle types, making it difficult to achieve precise scheduling and efficient traffic flow.

[0004] Therefore, to solve the above problems, a two-layer decision optimization method for multi-vehicle cooperative driving in intelligent connected mixed traffic environments is needed, which can achieve adaptive modeling and real-time optimization of multi-vehicle interaction behavior, and improve traffic efficiency and overall traffic stability under complex road conditions. Summary of the Invention

[0005] In view of this, the purpose of this invention is to overcome the deficiencies in the prior art and provide a two-layer decision optimization method for multi-vehicle cooperative driving in intelligent connected mixed traffic environments, which can realize adaptive modeling and real-time optimization of multi-vehicle interaction behavior, and improve traffic efficiency and overall traffic stability under complex road conditions.

[0006] The present invention provides a two-layer decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment, comprising:

[0007] Adjust lane type allocation to select the target lane for the vehicle;

[0008] The vehicle's driving intentions are quantified, and behavior search and target area allocation are performed to obtain the vehicle's driving decisions.

[0009] Based on driving decisions, each intelligent connected vehicle can smoothly travel from its current location to its target area.

[0010] Furthermore, selecting a target lane for the vehicle includes:

[0011] Construct an objective function, which includes a lane flow balancing function and a lane priority selection function;

[0012] Set constraints for the objective function, solve the objective function under the constraints, and use the lane corresponding to the solved objective function as the target lane for the vehicle.

[0013] Furthermore, the lane flow equalization function is determined according to the following formula:

[0014] ;

[0015] in, Indicates time period Section Lane Traffic capacity.

[0016] Furthermore, the lane priority selection function is determined according to the following formula:

[0017] ;

[0018] ;

[0019] in, Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates time period Section Lane Is it a lane type? , The options are "No" and "Yes".

[0020] Furthermore, the constraints are determined according to the following formula:

[0021] ;

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] ;

[0034] Then we have:

[0035]

[0036]

[0037]

[0038]

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] in, Indicates the time period Section Vehicle type flow; Represents a constant; This represents the traffic flow in lane j of road segment i during time period t; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates the time period Section Vehicle type Select lane Traffic volume.

[0047] Furthermore, the vehicle's driving intentions are quantified, and behavioral searches and target area allocations are performed to obtain the vehicle's driving decisions, specifically including:

[0048] Based on the vehicle's actions, multiple potential target areas are set for the vehicle;

[0049] When multiple vehicles intend to enter the same target area, they are coordinated according to priority, with higher priority vehicles being assigned to the target area first.

[0050] Furthermore, coordination is achieved through prioritization, with higher-priority areas allocated first. Specifically, this includes:

[0051] Manually driven vehicles always have a higher priority than intelligent connected vehicles. When intelligent connected vehicles compete for the area, the priority of going straight is higher than the priority of changing lanes. When all intelligent connected vehicles are changing lanes, the priority of forced lane changing is higher than the priority of free lane changing.

[0052] When all connected vehicles are capable of free lane changing, the Monte Carlo tree search algorithm is used to determine the connected vehicles with higher priority:

[0053] Action space definition: Design a discretized action set for each vehicle. The action space is the Cartesian product of the actions of each vehicle. The number of action combinations is limited to a certain range through pruning strategy.

[0054] A search tree is gradually built through iteration. Each iteration includes four steps: selection, expansion, simulation, and backpropagation.

[0055] Selection: Starting from the root node, the optimal node is continuously selected based on the strategy formula until the selected node is a non-extendable node; the strategy formula is:

[0056] ;in, It is a child node The value score; This is the total number of times the search tree has been explored so far; It is a node The number of times it has been explored; 𝜔 is the weighting coefficient;

[0057] Extension: Randomly generate a new child node for the selected node and add the new child node to the search tree structure;

[0058] Simulation: Starting from the newly expanded nodes, a random sampling simulation strategy is executed to continuously generate new child nodes. That is, a vehicle is randomly selected from those not yet included in the current traffic sequence and added to the end of that traffic sequence until a complete traffic sequence is generated. The node value score is then updated according to the following formula:

[0059] ;

[0060] in, It is a node The value score; It is a quantitative assessment value; It is a node The maximum value score among all sibling nodes; It is a node The minimum value score among all sibling nodes;

[0061] Backpropagation: Transferring the new value score of a node It propagates upwards along the direction of its parent node until it reaches the root node;

[0062] During the above iteration process, the current optimal passage order is constantly updated. When the maximum number of search iterations is reached, the iteration process stops and returns the current optimal leaf node, which corresponds to the intelligent connected vehicle with high priority.

[0063] Furthermore, based on driving decisions, each intelligent connected vehicle can smoothly travel from its current location to the target area, specifically including:

[0064] A third-order Bezier curve is used to pre-generate a smooth trajectory from the current position to the target area;

[0065] Expected velocity is generated based on trajectory length and expected time window. :

[0066] ;in, This is the adjustment factor in speed planning. For the trajectory length, For the desired time window, The initial velocity;

[0067] A proportional controller is used to generate longitudinal acceleration. :

[0068] ;in, For the maximum permissible acceleration, This represents the gain coefficient of the proportional controller. For current intelligent connected vehicles speed;

[0069] Analyze whether the closest distance between the trajectory points of two intelligent connected vehicles is lower than a threshold. If so, it is determined to be a conflict, and the vehicle to avoid is determined by the set priority. Use a collision ellipse model to determine whether there is a trajectory conflict between the intelligent connected vehicle and the manually driven vehicle. If so, the intelligent connected vehicle reduces acceleration to avoid collision when it is in a low priority position.

[0070] Select a forward-looking point in the trajectory and construct a preview vector. :

[0071] in, This indicates the distance between the current position of the car's front and the forward viewpoint. ; The x-coordinate of the foreground viewpoint. For the current vehicle The x-coordinate;

[0072] Based on the bicycle model, the angle between the preview vector and the vehicle's current orientation is... Converted to steering angle :

[0073] ;in, The distance from the rear wheel of the car to the preview point; This refers to the vehicle's wheelbase.

[0074] acceleration and steering angle The signal is sent to the corresponding intelligent connected vehicle, which then performs the action according to the corresponding acceleration and steering angle.

[0075] The beneficial effects of this invention are as follows: The invention discloses a two-layer decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment. Based on a lane-level and vehicle-level hierarchical architecture, it integrates static structural features and dynamic traffic state information. Addressing the mixed traffic control needs in an intelligent connected environment, it constructs a computationally efficient, finely scheduled, and highly scalable cooperative optimization mechanism. This effectively solves the core problems of traditional single optimization levels, weak scheduling capabilities, and lack of modeling for mixed traffic structures, providing technical support for the large-scale deployment of intelligent connected vehicles. Attached Figure Description

[0076] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0077] Figure 1 This is a schematic diagram illustrating the principle of the multi-vehicle cooperative driving two-layer group decision-making framework of the present invention;

[0078] Figure 2 This is a schematic diagram defining the lane-level decision-making scenario of the present invention;

[0079] Figure 3 This is a schematic diagram illustrating the delineation of the vehicle driving target area according to the present invention;

[0080] Figure 4 This is a schematic diagram illustrating the high-risk area screening method of the present invention;

[0081] Figure 5 This is a schematic diagram illustrating the priority setting for right-of-way conflicts in this invention.

[0082] Figure 6 This is a schematic diagram of the vehicle optimal decision behavior search based on Monte Carlo tree search according to the present invention;

[0083] Figure 7 This is a schematic diagram of the multi-vehicle cooperative driving two-layer group decision-making framework of the present invention;

[0084] Figure 8 This is a schematic diagram illustrating the implementation principle of the two-level decision-making method of the present invention. Detailed Implementation

[0085] The present invention will be further described below with reference to the accompanying drawings, as shown in the figures:

[0086] This embodiment discloses a two-layer decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment, including the following steps:

[0087] S1. Adjust lane type assignment and select the target lane for the vehicle;

[0088] S2. Quantify the vehicle's driving intentions, perform behavior search and target area allocation, and obtain the vehicle's driving decisions;

[0089] S3. Based on driving decisions, each intelligent connected vehicle can smoothly travel from its current location to the target area.

[0090] This invention designs a two-layer group decision-making framework for multi-vehicle cooperative driving in intelligent connected mixed traffic environments. The entire architecture consists of two layers: the upper layer is lane management, and the lower layer is vehicle-level multi-vehicle decision-making, such as... Figure 1 As shown:

[0091] Upper layer (lane-level decision layer): The vehicle management layer primarily selects available lanes based on vehicle type (ICV / HDV, passenger car / light truck / bus / truck). Considering the difficulty in accurately quantifying the probability formulas for lane selection by various vehicle types, it is assumed that all vehicles prioritize the lane with the lowest traffic flow among the permitted lanes; the specific traffic flow formula is detailed in Part Two. After adjusting the lane type allocation, the permitted lane set and traffic flow distribution are recalculated, iterating until convergence.

[0092] Lower layer (vehicle-level decision layer): Based on the decision results of road segment lane type, combined with the vehicle status and driving intention of the intelligent connected vehicle, the vehicle-level multi-vehicle decision layer is divided into two parts:

[0093] The first part is vehicle driving behavior-level decision-making. The grid method is used to quantify the vehicle's driving intentions, thereby detecting conflicts and avoiding obstacles. Then, target area allocation is performed based on MCTS behavior search. The allocation mode refers to road right-of-way priority, and the driving decision for each vehicle is output.

[0094] The second part is the specific vehicle trajectory planning. Based on the driving decision set in the first part, a trajectory is generated for each ICV using a third-order Bezier curve. Then, the specific speed is planned and conflicts are corrected. Finally, the driving action of each vehicle is output based on the bicycle model.

[0095] This invention introduces a layered decoupling mechanism to combine lane type configuration with vehicle behavior planning, achieving coordinated behavior and optimal path allocation for multiple vehicles sharing road resources while ensuring traffic safety. Based on real-time traffic data and vehicle perception information, it dynamically generates lane priority schemes and vehicle behavior instructions, adapting to various traffic flow structures and environmental changes, significantly improving adaptive capabilities and decision-making efficiency in mixed traffic flows. It establishes a unified decision-making foundation and scheduling system for multi-vehicle cooperative driving, empowering intelligent connected transportation with higher collaborative scheduling capabilities, resource allocation flexibility, and multi-dimensional safety assurance mechanisms. This drives the evolution of intelligent transportation from "single-vehicle intelligence" to "swarm intelligence," promoting the transformation of transportation systems from static management to dynamic perception and intelligent response, ultimately achieving optimal coordination of multiple objectives: traffic efficiency, safety, and comfort.

[0096] In this embodiment, step S1 involves lane-level decision-making, such as... Figure 2 As shown, time period Road section ,Lane Vehicle type These are, respectively, HDV passenger cars, HDV light trucks, HDV buses, HDV trucks, ICV passenger cars, ICV light trucks, ICV buses, and ICV trucks. These are designated as dedicated lanes for ICVs, mixed lanes for ICVs and HDVs, and dedicated lanes for HDVs. These are respectively the passenger car lane, passenger car lane, passenger and freight car lane, and freight car lane.

[0097] The probability of different vehicle types choosing a particular lane is closely related to lane management strategies, the proportion of each vehicle type, whether the road segment includes exit or entrance ramps, driver driving preferences, lane speed limits, and vehicle performance, namely:

[0098] ;

[0099] In the formula: For the time period Section Vehicle type Select lane The probability of driving; For time period Section Lane Speed ​​limit; For time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; For time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; For all types of vehicles Percentage; For the time period Section Vehicle type Select lane Traffic flow; For the time period Section Vehicle type flow; For time period Section Lane Traffic; For time period Section Lane Traffic capacity; , For the set of permitted lane management policies;

[0100] The creation of vehicle road models needs to be reasonable, and road types can be distinguished in two modes:

[0101] 1. Based on the type of vehicle being driven, namely ICV and HDV, lanes can be divided into dedicated lanes for ICVs, dedicated lanes for HDVs, and mixed lanes. The lane layout is based on the fact that ICVs can obtain more information, react faster, and have a smoother overall speed trend, resulting in a faster lane flow. Therefore, they are placed in the fast lane, i.e., the leftmost lane. Mixed lanes can only be set up in the middle, while dedicated lanes for HDVs are on the rightmost side.

[0102] 2. Based on the original attributes of the vehicles, namely passenger cars, buses, passenger-freight vehicles and freight vehicles, the lane layout is the same as above. Passenger cars are the smallest and faster, so they should be located on the far right of the lane. Ordinary buses should be on the left of the passenger car lane, passenger-freight vehicles should be on the left of the ordinary bus lane, and freight vehicles should be on the far left, satisfying an inclusion relationship.

[0103] Although road types can be distinguished by two models, only one model can be determined for road type classification. The choice between the two models is based on comparing the final values ​​that minimize their maximum lane saturation. The vehicle saturation of both lane assignment strategies is calculated, and the lane assignment strategy for each road segment is calculated separately. However, if the current road segment's lane assignment mode is based on vehicle driving type, the difference in lane type between adjacent road segments must be controllable (e.g., avoiding abrupt changes from ICV lanes to HDV lanes). If the current road segment's lane assignment mode is based on the vehicle's original attribute, lane type changes between consecutive road segments must be smooth (e.g., avoiding abrupt changes from passenger vehicle lanes to freight vehicle lanes). If the road segment contains exit / entry ramps, vehicles in the inner lanes can change lanes to the outer lanes, allowing for dynamic adjustment of lane selection.

[0104] If the above conditions are met, a target lane will be selected for the vehicle, specifically including:

[0105] Construct an objective function, which includes a lane flow balancing function and a lane priority selection function;

[0106] Set constraints for the objective function, solve the objective function under the constraints, and use the lane corresponding to the solved objective function as the target lane for the vehicle.

[0107] The lane flow balancing function involves minimizing the maximum lane saturation value.

[0108] ;in, Indicates time period Section Lane The throughput capacity is set at 2000 pcu / h.

[0109] The lane priority selection function involves lane priority selection:

[0110] ;

[0111] ;

[0112] in, Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates time period Section Lane Is it a lane type? , The options are "No" and "Yes".

[0113] The constraints are determined according to the following formula:

[0114] ;

[0115] ;

[0116] ;

[0117] ;

[0118] ;

[0119] ;

[0120] ;

[0121] ;

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] ;

[0127] Then we have:

[0128]

[0129]

[0130]

[0131]

[0132] ;

[0133] ;

[0134] ;

[0135] ;

[0136] ;

[0137] ;

[0138] ;

[0139] in, Indicates the time period Section Vehicle type Flow rate, in pcu / h; This represents a constant and can take the value 10000. Indicates time period Section Lane The flow rate, in units of pcu / h; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates the time period Section Vehicle type Select lane Traffic volume. It is a 0-1 variable, if the road segment If the area is located 2km upstream to 0m downstream of the diversion zone and 0m upstream to 2km downstream of the merging zone, then... ,otherwise . This is an intermediate variable, in units of pcu / h.

[0140] It should be noted that weighting coefficients can also be set for the lane flow balancing function and the lane priority selection function in the objective function. as well as The value can be 1000 or 1.

[0141] In this embodiment, step S2 involves vehicle behavior search for vehicle-level behavior decision-making, quantifying the vehicle's driving intention, performing behavior search and target area allocation, and obtaining the vehicle's driving decision, specifically including:

[0142] The Monte Carlo Tree Search (MCTS) algorithm is used to construct a behavior search tree, mapping the executable actions of intelligent connected vehicles (such as left turn, straight ahead, right turn, acceleration, and deceleration) to... Figure 3 The numbered areas (0-5 as shown in the figure) are used to define the driving behaviors of a vehicle. These behaviors can be primarily categorized into two main types: following and lane changing. Further subdivisions include acceleration and deceleration during following, left and right turns during lane changing, and overtaking on the left or right. To quantitatively and comprehensively evaluate all driving behaviors, six potential target areas are defined, such as... Figure 3 As shown. First, high-risk routes with high-risk areas (such as construction or blast signs) or traffic restrictions (such as no-entry signs or traffic lights) are filtered out, forming a feasibility vector. Figure 4 As shown.

[0143] like Figure 5 As shown, when multiple vehicles intend to enter the same target area, coordination is achieved through priority, with higher priority vehicles being allocated the target area first. Specifically, this includes:

[0144] Manually driven vehicles always have a higher priority than intelligent connected vehicles. When intelligent connected vehicles compete for the area, the priority of going straight is higher than the priority of changing lanes. When all intelligent connected vehicles are changing lanes, the priority of forced lane changing is higher than the priority of free lane changing.

[0145] When all connected vehicles are capable of free lane changing, the Monte Carlo tree search algorithm is used to determine the connected vehicles with higher priority:

[0146] In the Monte Carlo tree search algorithm, each child node in the search tree has a score representing its own value. In the application of right-of-way allocation problems, the evaluation strategy for the child node value scores is often related to the corresponding objective function. Based on the different child node value scores, the Monte Carlo tree search strategy will determine which branches of the search tree are expanded first.

[0147] Action space definition, designing a discretized action set for each vehicle:

[0148] The motion space of a group is the Cartesian product of the motions of each vehicle. = … By using pruning strategies, the number of action combinations is limited to a certain limit. =Within 125 (when m=3);

[0149] A search tree is gradually built through iteration. Each iteration includes four steps: selection, expansion, simulation, and backpropagation. Figure 6 As shown:

[0150] Selection: Starting from the root node, the optimal node is continuously selected based on the strategy formula until the selected node is a non-extendable node; the strategy formula is:

[0151] ;in, It is a child node The value score; This is the total number of times the search tree has been explored so far; It is a node The number of times it has been explored; 𝜔 is the weighting coefficient, which can take values ​​of ;

[0152] In the iterative process of Monte Carlo tree search, an expandable node is a node that has unexplored child nodes and is not a leaf node itself. In the "selection" step, the child node with the highest value score is selected as the next node to expand. The first term in the strategy formula consists only of the node's value score, aiming to encourage the selection of the child node with the highest current value score during the search. The second term in the strategy formula considers both the current node's exploration count and the total number of explorations, assigning a higher value score to nodes that have been explored less frequently, thus encouraging the exploration of more new nodes during the search.

[0153] The first aspect of this strategy tends to be a "conservative" approach, which is to directly select the node with the best current value score. The second aspect tends to be a "risky" approach, which attempts to explore as many new nodes as possible to find the global optimum. The weight 𝜔 is used to balance the two opposing search tendencies.

[0154] Of course, three improvements can be introduced based on the above strategy:

[0155] (1) Dynamic depth limit: The search depth is dynamically adjusted according to the urgency of the group:

[0156] ;

[0157] (2) Heuristic reward function:

[0158] Driving benefits are related not only to the vehicle's state but also to the degree of conflict between vehicles. Therefore, safety, speed, and comfort indicators are used to reflect the reward function of the decision-maker.

[0159] a. Security

[0160] Safety is the primary indicator for avoiding collisions during traffic at intersections. Assuming... This indicates the time difference between the arrival of the vehicles at the point of conflict. The smaller the value, the greater the intensity of the vehicle collision and the lower the driving benefit. The safety index is calculated as follows:

[0161]

[0162] In the formula: , They are respectively Time of the first Target vehicle and the The time it takes for the vehicles involved in the conflict to arrive at the point of conflict.

[0163] b. Speed

[0164] Speed ​​refers to the expected speed at which vehicles can pass through intersections or merging zones. Speed ​​is reflected in the speed indicator through changes in speed.

[0165]

[0166] In the formula: Δ Δ For vehicles ,vehicle exist Changes in velocity over time.

[0167] c. Comfort

[0168] Comfort primarily considers changes in the vehicle's longitudinal acceleration. Excessive changes in acceleration can lead to a decrease in comfort. Therefore, the amount of acceleration change is chosen as the indicator to reflect comfort.

[0169]

[0170] In the formula: Δ Δ For vehicles ,vehicle From the Time to the The change in acceleration at any given moment includes all actions and behaviors within the conflict resolution strategy set.

[0171] Extension: Randomly generate a new child node for the selected node and add the new child node to the search tree structure;

[0172] Simulation: Starting from the newly expanded node, a random sampling simulation strategy is executed to continuously generate new child nodes. Specifically, a vehicle is randomly selected from those not yet included in the current traffic sequence and added to the end of that traffic sequence until a complete traffic sequence is generated. Figure 6 Taking the CB node in the search process as an example, after executing the random sampling simulation strategy, a new child node (such as CBE) will be randomly generated. Then, the random sampling simulation strategy will continue to be executed on the CBE node, and so on, until a leaf node (such as CBEDA) is generated.

[0173] After obtaining the complete passage order, update the node value score according to the following formula:

[0174] ;

[0175] in, It is a node The value score; It is a quantitative evaluation value, which can take values ​​of [-0.33, 66.7]. It is a node The maximum value score among all sibling nodes; It is a node The minimum value score among all sibling nodes;

[0176] Backpropagation: Transferring the new value score of a node It propagates upwards along the direction of its parent node until it reaches the root node;

[0177] During the above iteration process, the current optimal passage order is constantly updated. When the maximum number of search iterations is reached, the iteration process stops and returns the current optimal leaf node, which corresponds to the intelligent connected vehicle with high priority.

[0178] In this embodiment, step S3 involves vehicle trajectory planning for vehicle-level behavioral decision-making. Based on driving decisions, it enables each intelligent connected vehicle to smoothly travel from its current location to the target area, specifically including:

[0179] A third-order Bezier curve is used to pre-generate a smooth trajectory from the current location to the target area. The third-order Bezier curve offers good smoothness and computational efficiency, making it suitable for path planning in traffic environments. The trajectory is defined by four control points:

[0180] ;

[0181] ;

[0182] Control point settings: Point 1 is the current vehicle position; Point 4 is the center of the target area assigned by MCTS; intermediate control points are offset by a fixed ratio to control the curve shape. The trajectory is discretized into 101 points, forming a trajectory point set. ;

[0183] Expected velocity is generated based on trajectory length and expected time window. :

[0184] ;in, This is an adjustment factor in speed planning, which can be set according to actual operating conditions, for example, a value of 1.0; For the trajectory length, For the desired time window, The initial velocity;

[0185] A proportional controller is used to generate longitudinal acceleration. :

[0186] ;in, For the maximum permissible acceleration, This is the gain coefficient of the proportional controller. Similarly, this gain coefficient can be set according to the actual operating conditions, for example, a value of 0.5. For the current vehicle speed;

[0187] Analyze whether the closest distance between the trajectory points of two intelligent connected vehicles is lower than a threshold. If so, it is determined to be a conflict, and the vehicle to avoid is determined by the set priority.

[0188] A collision ellipse model is used to determine whether there is a trajectory conflict between the intelligent connected vehicle and the manually driven vehicle. If so, the intelligent connected vehicle reduces acceleration to avoid a collision when it is in a low-priority position.

[0189] ; For the maximum permissible deceleration, For intelligent connected vehicles The current speed, For manually driven vehicles The current speed. Considering the uncertainty of the trajectory of manually driven vehicles, a collision ellipse model is used to determine potential conflicts:

[0190] ;in, .

[0191] Select a forward-looking point in the trajectory and construct a preview vector. :

[0192] in, This indicates the distance between the current position of the car's front and the forward viewpoint. ; The x-coordinate of the foreground viewpoint is... For the current vehicle The x-coordinates mentioned above belong to the global map coordinate system or the map Cartesian coordinate system, which will not be elaborated further here.

[0193] Based on the bicycle model, the angle between the preview vector and the vehicle's current orientation is... Converted to steering angle :

[0194] ; ;in, The distance from the rear wheel of the car to the preview point; This refers to the vehicle's wheelbase.

[0195] acceleration (Longitudinal control) and steering angle (Lateral control) is sent in the form of instructions to the vehicle actuators of the corresponding intelligent connected vehicle, which then performs actions according to the corresponding acceleration and steering angle.

[0196] This invention can be widely applied to multi-vehicle cooperative control scenarios in intelligent connected mixed traffic environments, and is particularly suitable for complex road environments such as highway mainline merging, urban expressway intersections, and temporary construction lane change zones. By introducing a roadside cooperative control unit to achieve global perception and intelligent guidance of the status of multiple vehicles, it realizes efficient scheduling and dynamic right-of-way allocation for coordinated vehicle behavior, effectively improving the traffic efficiency and operational safety of key road sections.

[0197] The proposed two-layer group decision-making control method fully leverages the advantages of lane-level and vehicle-level collaborative control, significantly enhancing its adaptability to traffic flow changes. During multi-vehicle collaborative scheduling, high-frequency V2V / V2I communication and game-theoretic optimization algorithms enable real-time adjustments to vehicle paths, speeds, and timing, effectively mitigating traffic conflicts in weaving areas and reducing potential traffic risks arising from inconsistent vehicle behavior. Simultaneously, this method improves traffic efficiency while also considering driving comfort and operational stability, providing reliable control support for high-density traffic scenarios.

[0198] This invention possesses excellent scalability and portability, and has broad application potential in future intelligent transportation systems. Its control mechanism can be seamlessly embedded into existing intelligent transportation infrastructure, driving the upgrade of urban road traffic systems from traditional static management to a dynamic, adaptive, and collaborative control mode. Overall, this invention not only provides key technical support for the large-scale deployment of intelligent connected vehicles, but also lays a solid foundation for achieving a more efficient, safe, and green intelligent transportation ecosystem, demonstrating significant engineering application prospects.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A two-level decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment, characterized in that: include: Adjust lane type allocation to select the target lane for the vehicle; Selecting a target lane for a vehicle includes: Construct an objective function, which includes a lane flow balancing function and a lane priority selection function; Set constraints for the objective function, solve the objective function under the constraints, and take the lane corresponding to the solved objective function as the target lane for the vehicle. The lane flow equalization function is determined according to the following formula: ； in, Indicates time period Section Lane Traffic capacity; road section ,Lane ; This represents the traffic flow in lane j of road segment i during time period t; The lane priority selection function is determined according to the following formula: ； ； in, Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; These are designated as dedicated lanes for ICVs, mixed lanes for ICVs and HDVs, and dedicated lanes for HDVs. These are respectively the passenger car lane, the bus lane, the passenger and freight vehicle lane, and the freight vehicle lane; The vehicle's driving intentions are quantified, and behavioral searches and target area allocations are performed to obtain the vehicle's driving decisions, specifically including: Based on the vehicle's actions, multiple potential target areas are set for the vehicle; When multiple vehicles intend to enter the same target area, they are coordinated according to priority, with higher priority vehicles being assigned to the target area first. Based on driving decisions, each intelligent connected vehicle can smoothly travel from its current location to its target area.

2. The dual-level decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment according to claim 1, characterized in that: The constraints are determined according to the following formula: ； ； ； ； ； ； ； ； ； ； ； ； ； Then we have: ； ； ； ； ； ； ； in, Indicates the time period Section Vehicle type flow; Represents a constant; This represents the traffic flow in lane j of road segment i during time period t; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates time period Section Lane Is it a lane type? , They are respectively "no" and "yes"; Indicates the time period Section Vehicle type Select lane Traffic volume; vehicle type These are HDV passenger cars, HDV light trucks, HDV buses, HDV trucks, ICV passenger cars, ICV light trucks, ICV buses, and ICV trucks.

3. The dual-level decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment according to claim 1, characterized in that: Coordination is based on priorities, with higher-priority areas allocated first. Specifically, this includes: Manually driven vehicles always have a higher priority than intelligent connected vehicles. When intelligent connected vehicles compete for the area, the priority of going straight is higher than the priority of changing lanes. When all intelligent connected vehicles are changing lanes, the priority of forced lane changing is higher than the priority of free lane changing. When all connected vehicles are capable of free lane changing, the Monte Carlo tree search algorithm is used to determine the connected vehicles with higher priority: Action space definition: Design a discretized action set for each vehicle. The action space is the Cartesian product of the actions of each vehicle. The number of action combinations is limited to a certain range through pruning strategy. A search tree is gradually built through iteration. Each iteration includes four steps: selection, expansion, simulation, and backpropagation. Selection: Starting from the root node, the optimal node is continuously selected based on the strategy formula until the selected node is a non-extendable node; the strategy formula is: ;in, It is a child node The value score; This is the total number of times the search tree has been explored so far; It is a node The number of times it has been explored; 𝜔 is the weighting coefficient; Extension: Randomly generate a new child node for the selected node and add the new child node to the search tree structure; Simulation: Starting from the newly expanded nodes, a random sampling simulation strategy is executed to continuously generate new child nodes. That is, a vehicle is randomly selected from those not yet included in the current traffic sequence and added to the end of that traffic sequence until a complete traffic sequence is generated. The node value score is then updated according to the following formula: ； in, It is a node The value score; It is a quantitative assessment value; It is a node The maximum value score among all sibling nodes; It is a node The minimum value score among all sibling nodes; Backpropagation: Transferring the new value score of a node It propagates upwards along the direction of its parent node until it reaches the root node; During the above iteration process, the current optimal passage order is constantly updated. When the maximum number of search iterations is reached, the iteration process stops and returns the current optimal leaf node, which corresponds to the intelligent connected vehicle with high priority.

4. The dual-level decision optimization method for multi-vehicle cooperative driving in an intelligent connected mixed traffic environment according to claim 1, characterized in that: Based on driving decisions, each intelligent connected vehicle can smoothly travel from its current location to its target area, specifically including: A third-order Bezier curve is used to pre-generate a smooth trajectory from the current position to the target area; Expected velocity is generated based on trajectory length and expected time window. : ;in, This is the adjustment factor in speed planning. For the trajectory length, For the desired time window, The initial velocity; A proportional controller is used to generate longitudinal acceleration. : ;in, For the maximum permissible acceleration, This represents the gain coefficient of the proportional controller. For current intelligent connected vehicles speed; Analyze whether the closest distance between the trajectory points of two intelligent connected vehicles is lower than a threshold. If so, it is determined to be a conflict, and the vehicle to avoid is determined by the set priority. Use a collision ellipse model to determine whether there is a trajectory conflict between the intelligent connected vehicle and the manually driven vehicle. If so, the intelligent connected vehicle reduces acceleration to avoid collision when it is in a low priority position. Select a forward-looking point in the trajectory and construct a preview vector. : in, This indicates the distance between the current position of the car's front and the forward viewpoint. ; The x-coordinate of the foreground viewpoint. For the current vehicle The x-coordinate; Based on the bicycle model, the angle between the preview vector and the vehicle's current orientation is... Converted to steering angle : ;in, The distance from the rear wheel of the car to the preview point; This refers to the vehicle's wheelbase. acceleration and steering angle The signal is sent to the corresponding intelligent connected vehicle, which then performs the action according to the corresponding acceleration and steering angle.

Images