A mixed traffic flow signal-free cooperative scheduling method, system, device and medium

Abstract

The application provides a mixed traffic flow signal-free cooperative scheduling method, system, device and medium, comprising: acquiring traffic sensing data, generating predicted driving trajectories of all vehicles according to the traffic sensing data, then determining vehicles that may have conflicts according to the predicted driving trajectories to obtain a vehicle cluster; constructing a trajectory optimization model, optimizing the predicted driving trajectories of all vehicles in the vehicle cluster using the trajectory optimization model, and extracting a plurality of conflict vehicle pairs that must have conflicts; setting a passing priority for vehicles in each conflict vehicle pair according to the types of the vehicles and traffic rules; reconstructing the trajectory optimization model, then optimizing the predicted driving trajectories of all vehicles in the vehicle cluster using the reconstructed trajectory optimization model to obtain an optimized trajectory corresponding to each vehicle, and then sending the optimized trajectory to the corresponding vehicle. The application solves the problems of decision lag and low decision accuracy of the automatic driving decision method in the prior art.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation systems, and in particular to a signalless collaborative scheduling method, system, device, and medium for mixed traffic flow. Background Technology

[0002] In intelligent connected vehicle environments, left turns at unsignalized intersections present a typical high-complexity, high-stakes scenario for autonomous vehicles. In mixed traffic flows where intelligent connected vehicles and human-driven vehicles coexist, the behavioral patterns of traffic participants are complex, variable, and difficult to predict. Traditional autonomous driving decision-making methods generally rely on single-vehicle perception, which suffers from limited field of vision, blind spots, and data redundancy. They fail to acquire crucial information about areas with obstructed vision, easily leading to delayed or erroneous decisions and posing safety hazards.

[0003] Furthermore, existing decision-making models have limitations when dealing with fuzzy traffic rules (such as "ensure safety" and "do not obstruct others") and dynamic priorities. Many models apply the same weights to all rules, failing to dynamically adjust their importance based on the real-time traffic context, leading to decreased decision-making accuracy in complex scenarios. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a signalless cooperative scheduling method for mixed traffic flow, which solves the problems of decision lag and low accuracy in existing autonomous driving decision-making methods.

[0005] According to an embodiment of the present invention, a signalless cooperative scheduling method for mixed traffic flow includes:

[0006] Acquire traffic perception data, generate predicted driving trajectories for all vehicles based on the traffic perception data, and then determine vehicles that may collide based on the predicted driving trajectories to obtain a vehicle cluster.

[0007] Construct a trajectory optimization model and constraints. Based on the constraints, use the trajectory optimization model to optimize the predicted driving trajectory of all vehicles in the vehicle cluster and extract multiple conflict vehicle pairs that are bound to collide.

[0008] Calculate the conflict risk index for each conflicting vehicle pair, set the passage priority for each vehicle in the conflicting vehicle pair based on the conflict risk index, the type of vehicle in the conflicting vehicle pair and traffic rules, and then combine the vehicles of all conflicting vehicle pairs into a passage sequence.

[0009] The trajectory optimization model is reconstructed based on the vehicle type and traffic rules of the conflicting vehicles. Then, the reconstructed trajectory optimization model is used to optimize the predicted driving trajectory of all vehicles in the vehicle cluster, so as to obtain the optimized trajectory for each vehicle. Finally, the optimized trajectory is sent to the corresponding vehicle.

[0010] Preferably, the vehicle includes intelligent connected vehicles and manually driven vehicles;

[0011] Before generating the predicted driving trajectory, the speed of all intelligent connected vehicles is adjusted based on traffic perception data to ensure that they drive at the optimal speed.

[0012] For intelligent connected vehicles, a position change function is constructed based on traffic perception data, and a first predicted driving trajectory is generated based on the position change function. For manually driven vehicles, multiple second predicted driving trajectories and corresponding confidence scores are generated based on the GAT learning model using traffic perception data. Then, all second predicted driving trajectories with confidence scores less than the confidence score threshold are eliminated.

[0013] Preferably, the method for determining vehicles that may collide based on predicted driving trajectories to obtain a vehicle cluster is as follows:

[0014] The intersection points between all trajectories are determined based on the predicted driving trajectories of all vehicles, and multiple conflict zones are generated with the intersection points as the center.

[0015] Calculate the predicted conflict time when all intelligent connected vehicles reach any conflict area while traveling along their corresponding predicted trajectories.

[0016] Create a vehicle cluster. For any two intelligent connected vehicles, if the minimum distance between their corresponding predicted trajectories is less than the safe distance and the difference in conflict time is less than the safe time, then add the two intelligent connected vehicles to the vehicle cluster.

[0017] For any intelligent connected vehicle and any manually driven vehicle, if the first predicted driving trajectory of the intelligent connected vehicle and any second predicted driving trajectory of the manually driven vehicle intersect in the same conflict area, then the intelligent connected vehicle and the manually driven vehicle are added to the vehicle cluster.

[0018] Preferably, the constraints include vehicle dynamics constraints and collision-free constraints;

[0019] The collision-free constraint between two intelligent connected vehicles is:

[0020]

[0021] in, Intelligent connected vehicles and intelligent connected vehicles At any moment The position vector, , For intelligent connected vehicles and intelligent connected vehicles The safety radius, The set consists of all intelligent connected vehicles in the vehicle cluster, and Tf is the trajectory duration;

[0022] The collision-free constraint between intelligent connected vehicles and manually driven vehicles is:

[0023]

[0024] In the formula: For intelligent connected vehicles At any moment The position vector, For manually driven vehicles The The predicted driving trajectory at time... The position vector, For intelligent connected vehicles and manually driven vehicles The safety radius, For manually driven vehicles The set of all predicted driving trajectories For manually driven vehicles The The confidence level of the predicted driving trajectory, This is the confidence threshold.

[0025] Preferably, the method for setting traffic priority for each vehicle in a conflicting vehicle pair based on conflict risk indicators, the types of vehicles in the conflicting vehicle pair, and traffic rules includes:

[0026] Based on the magnitude of the conflict risk index, all conflict vehicle pairs are sorted in descending order to obtain the conflict sequence to be processed.

[0027] Create a yield priority list, iterate through all the first conflict vehicle pairs in the pending conflict sequence, and in response to the conflict vehicle pair being intelligent connected vehicle i and human-driven vehicle m, add intelligent connected vehicle i to the yield priority list, and then remove all conflict vehicle pairs related to intelligent connected vehicle i from the pending conflict sequence.

[0028] In response to a conflicting vehicle pair being intelligent connected vehicle i and intelligent connected vehicle j, intelligent connected vehicle i or intelligent connected vehicle j is added to the yield priority list according to preset traffic rules, and then all conflicting vehicle pairs related to intelligent connected vehicle i or intelligent connected vehicle j are removed from the pending conflict sequence.

[0029] Preferably, when both conflicting vehicle pairs are intelligent connected vehicles and intelligent connected vehicle i needs to be added to the yield priority list according to preset traffic rules, if the number of manually driven vehicles associated with intelligent connected vehicle i in the conflict sequence to be processed is greater than the number of manually driven vehicles associated with intelligent connected vehicle j, then intelligent connected vehicle j is added to the yield priority list, and then all conflicting vehicle pairs associated with intelligent connected vehicle j are deleted from the conflict sequence to be processed.

[0030] Preferably, the reconstructed trajectory optimization model is as follows:

[0031]

[0032] in, The trajectory optimization model before reconstruction As a penalty weight, For the yield priority list, For intelligent connected vehicles At any moment speed, The target speed for yielding to intelligent connected vehicle i.

[0033] On the other hand, according to embodiments of the present invention, a mixed traffic flow signalless cooperative scheduling system is also provided, which uses the above-described mixed traffic flow signalless cooperative scheduling method, including:

[0034] A perception module, which is used to acquire traffic perception data and send optimized trajectories to the corresponding vehicles;

[0035] The trajectory generation module is used to generate predicted driving trajectories for all vehicles based on traffic perception data, and then determine vehicles that may conflict based on the predicted driving trajectories to obtain a vehicle cluster.

[0036] A model building module, which is used to build and reconstruct trajectory optimization models;

[0037] A condition preset module, which is used to set constraints and penalty rules;

[0038] The trajectory optimization module is used to optimize the predicted driving trajectory of all vehicles in the vehicle cluster using the trajectory optimization model, extract multiple conflicting vehicle pairs that are bound to collide and set the passage priority for the vehicles in each conflicting vehicle pair, then combine the vehicles of all conflicting vehicle pairs into a passage sequence, and use the reconstructed trajectory optimization model to optimize the predicted driving trajectory of all vehicles in the vehicle cluster to obtain the optimized trajectory for each vehicle.

[0039] On the other hand, according to an embodiment of the present invention, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the above-described method for unsignaled coordinated scheduling of mixed traffic flows.

[0040] On the other hand, according to an embodiment of the present invention, a computer storage medium is also provided, storing a computer program, which, when executed by a processor, causes the processor to perform the above-described signalless coordinated scheduling method for mixed traffic flow.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] This invention generates a first predicted driving trajectory and a second predicted driving trajectory for controllable intelligent connected vehicles and uncontrollable manually driven vehicles based on traffic perception data. Then, the first predicted driving trajectory is adjusted according to preset penalty rules and the second predicted driving trajectory, thereby making the collaborative decision-making of intelligent connected vehicles more robust and safe, and improving decision-making accuracy. At the same time, when adjusting the first predicted driving trajectory, a trajectory optimization model is used to decompose the large-scale collaborative problem into dynamic clusters, and a lightweight iterative optimization method is adopted to ensure that the system can make decisions within milliseconds, meeting real-time requirements. Attached Figure Description

[0043] Figure 1 This is a diagram of a signalless coordinated scheduling method for mixed traffic flow according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of a traffic scenario according to an embodiment of the present invention.

[0045] Figure 3 This is a schematic diagram of a conflict area in a traffic scenario according to an embodiment of the present invention. Detailed Implementation

[0046] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0047] like Figure 1 As shown in the figure, this invention proposes a signalless cooperative scheduling method for mixed traffic flow, including:

[0048] Acquire traffic perception data, generate predicted driving trajectories for all vehicles based on the traffic perception data, and then determine vehicles that may collide based on the predicted driving trajectories to obtain a vehicle cluster.

[0049] The traffic scenarios of this invention are as follows: Figure 2 As shown, the traffic scenario at this intersection includes the following three types of areas:

[0050] (1) Free driving zone: Located outside the intersection, vehicles drive autonomously in this area according to normal traffic rules without being directly interfered with by the roadside cooperative control unit (RCCU).

[0051] (2) Core Passage Area: This is the core area where multiple vehicle trajectories intersect and there is a potential risk of collision. Within this area, the RCCU will perform precise conflict assessment and resolution, and issue accurate collaborative trajectories to ensure the safe and efficient passage of the intelligent connected vehicle cluster.

[0052] (3) Speed ​​adjustment preparation area: This area is a buffer zone for intelligent connected vehicles to enter the cooperative control range. The RCCU continuously monitors the intelligent connected vehicles in this area. When an intelligent connected vehicle enters this area, the RCCU will calculate and issue a guidance speed command in real time based on the current traffic load of the core passage area, the target path of the intelligent connected vehicle, and the expected time window for entering the core passage area.

[0053] The core function of the speed control zone is to smoothly adjust the speed of connected vehicles (e.g., accelerating to fill a time gap or decelerating to avoid unnecessary stops in the core area), enabling them to reach the entrance of the core traffic zone at optimal speed and comfort within a predetermined time window. This pre-coordinated speed significantly reduces the burden of complex trajectory adjustments and planning within the core traffic zone, improving overall traffic efficiency and driving comfort. The guiding speed command here is sent to the connected vehicle via V2X, and the vehicle's onboard unit performs simple speed tracking control.

[0054] In this traffic scenario, vehicles are generally divided into intelligent connected vehicles and human-driven vehicles. Intelligent connected vehicles have V2X communication, high-precision positioning, and autonomous driving capabilities. They can receive trajectories and instructions issued by the RCCU and execute them accurately. Human-driven vehicles are driven by humans, and their behavior is uncertain. Human-driven vehicles are regarded as moving obstacles in the scenario or external intelligent agents whose behavior needs to be predicted. The RCCU cannot directly control them, but it will predict their trajectories for the intelligent connected vehicles to make decisions.

[0055] In the aforementioned traffic scenario, this invention acquires traffic perception data from the entire traffic scenario. This traffic perception data includes high-precision perception data from RCCU (including radar, cameras, etc., used for detection of manually driven vehicles and non-connected intelligent connected vehicles), V2X communication data of intelligent connected vehicles (including location, speed, heading, etc.), and high-precision maps (providing road network topology, lane lines, and geometric information of conflict areas).

[0056] The RCCU fuses all traffic perception data in a unified global coordinate system, using fusion algorithms such as the Extended Kalman Filter (EKF) or the Unscented Kalman Filter (UKF) to generate a comprehensive state list containing all traffic participants (intelligent connected vehicles and human-driven vehicles). The goal of fusion is to improve perception accuracy and robustness, especially to compensate for the field-of-view limitations of onboard sensors.

[0057] Each participant state vector Defined as: ,in, It is the precise location. It's speed. It is the heading angle. It's angular velocity. It is labeled as either an intelligent connected vehicle or a manually driven vehicle. This is the vehicle's dimensions (length and width). It is its destination. It is the vehicle's intended path.

[0058] Before generating a predicted driving trajectory for a vehicle, it is necessary to analyze the intelligent connected vehicles within the speed control zone. Generate a boot speed This allows them to reach the core access area entrance within the desired time window.

[0059] RCCU first calculates the intelligent connected vehicle To the entrance of the core passage area distance Based on current traffic conditions and expected entry times into the core traffic area. Calculate the required average speed: .

[0060] The constraints that the above parameters need to satisfy are: the generated Vehicle dynamics and comfort constraints must be met, i.e. ,in Maximum acceleration for comfort (default: 2.5 m / s²) 2 ).

[0061] Then the RCCU will guide the speed. The information is sent via V2X communication to the corresponding connected vehicle, which then performs simple speed tracking control on its onboard unit to adjust to the optimal speed before entering the core traffic area.

[0062] Next, trajectory prediction is performed. Since the traffic scenario includes controllable intelligent connected vehicles and uncontrollable manually driven vehicles, trajectory prediction needs to be performed separately for each.

[0063] (1) Trajectory prediction of intelligent connected vehicles

[0064] For intelligent connected vehicles located in the speed adjustment preparation zone or core traffic zone Its first predicted driving trajectory Based on its current state Extrapolating from the predetermined optimal travel path and the guiding speed issued by the RCCU, the predicted trajectory of the intelligent connected vehicle is determined by constructing its position change function. This ensures that the predicted trajectory of the intelligent connected vehicle takes into account its future control objectives.

[0065]

[0066]

[0067] In the formula, For intelligent connected vehicles In the future The predicted location, , For intelligent connected vehicles At the current moment ( The initial position is obtained from traffic sensing data. For intelligent connected vehicles The initial velocity at the current moment, For intelligent connected vehicles At the initial heading angle at the current moment, The vehicle's acceleration components in the X and Y directions can be initially assumed to be 0 or based on the current acceleration, but will be adjusted after receiving the guide velocity to reach the guide velocity at the end of the prediction time domain. To predict time variables in the time domain, the range of values ​​is typically [range missing]. .

[0068] Then, the first predicted driving trajectory is generated based on the aforementioned position change function. .

[0069] (2) Prediction of the trajectory of manually driven vehicles

[0070] The behavior of manually driven vehicles is uncertain, and their second predicted trajectory... Rather than a simple extrapolation, this invention employs an existing vehicle trajectory prediction model based on Transformer and graph attention networks. This model extracts data features from the historical trajectory of the manually driven vehicle, surrounding traffic environment information (including the intentions of other intelligent connected vehicles and the manually driven vehicle), and the predicted trajectory of the intelligent connected vehicle. It then performs multimodal fusion on these data features, transforms the fused features into a feature map with multiple nodes, uses GAT interactive encoding to enhance the features of each node in the feature map, and finally uses a Transformer decoder for autoregressive prediction, generating a set of multiple predicted driving trajectories. Each second predicted driving trajectory Includes a confidence score Then, all second predicted driving trajectories with confidence scores less than the confidence score threshold are removed. This model predicts complex behaviors of manually driven vehicles, such as acceleration, deceleration, lane changing, and steering, taking into account the cognitive reaction time of the driver of the manually driven vehicle (default value: 1.2 seconds) and different driving styles.

[0071] according to Figure 3 All the predicted driving trajectories of all vehicles will intersect to some extent. The vehicles corresponding to these intersecting predicted driving trajectories are considered to be likely to collide. Since vehicles have volume and are not collision-free points, this invention defines the area around the intersection of the predicted driving trajectories as the conflict zone.

[0072] The collection of all conflict zones It is the set of all potential collision points in the core traffic area of ​​the intersection. This invention uses an elliptical model to perform precise mathematical modeling of each conflict area.

[0073] For each pair of predicted driving trajectories that may conflict, their intersection region is abstracted as an ellipse. Its mathematical expression is:

[0074]

[0075] The geometric intersection points of two predicted trajectories that intersect are extracted from the high-precision map. The centroids of these intersection points are determined as the centers of the ellipse. .

[0076] Semi-major axis ( The length of the intersection is determined by the length of the intersection area:

[0077]

[0078] in , The effective length of the two paths that constitute the conflict zone within the intersection area is obtained from map data. For example, for a turning path, its effective length is the length of the turning arc with potential interactive conflict, and for a straight path, it is its straight length within the conflict zone. This is a longitudinal safety margin used to ensure that the ellipse can cover the actual dimensions of the vehicle and longitudinal movement errors; the default value is 2 meters.

[0079] Semi-short axis ( The length of the intersection is determined by the width of the two paths in the intersection area:

[0080]

[0081] in , The effective width of the two paths that constitute the conflict zone is obtained from map data; for example, the width of a single lane is typically 3.5 meters. This is a lateral safety margin used to ensure that the ellipse can cover the vehicle's lateral dimensions and lateral movement errors; the default value is 1 meter.

[0082] For any two vehicles and Its predicted trajectory and (or In a certain elliptical conflict region Arrival time window within and If overlap exists, it is considered a potential collision risk. .

[0083] Potential conflict assessment for intelligent connected vehicles: Calculate the time it takes for all intelligent connected vehicles to reach any conflict area while traveling along their predicted trajectories. If the intelligent connected vehicles... and intelligent connected vehicles The minimum distance of the first predicted driving trajectory within the core traffic area is less than the set safety threshold. (The default value is 3 meters, which can be adjusted according to the actual application scenario and vehicle size), and the time windows for its arrival at any conflict zone overlap (i.e., the difference in conflict time). If they are considered to be in potential conflict, then they are considered to be in conflict.

[0084] Conflict detection between manually driven vehicles and intelligent connected vehicles: The RCCU will check all manually driven vehicles. Determine if a vehicle is driven manually. Will all second predicted driving trajectories be related to intelligent connected vehicles? or Their trajectories intersected within the same conflict zone.

[0085] Ultimately, the RCCU organizes all potentially conflicting intelligent connected vehicles and related manually driven vehicles into a vehicle cluster and constructs an independent cluster state diagram for it.

[0086] Construct a trajectory optimization model and constraints. Based on the constraints, use the trajectory optimization model to optimize the predicted driving trajectory of all vehicles in the vehicle cluster and extract multiple conflict vehicle pairs that are bound to collide.

[0087] Taking into account the overall traffic efficiency and smoothness, this invention constructs a trajectory optimization model. Since manually driven vehicles are uncontrollable, this trajectory optimization model can only consider the total cost of all intelligent connected vehicles. as follows:

[0088]

[0089] in, Let be the total cost function, used to measure the quality of the optimized predicted driving trajectory, and the goal is to minimize it. It is a set consisting of all intelligent connected vehicles in the vehicle cluster. The planning time domain, i.e., the duration of the trajectory, is typically taken as a value within... In this method, 5 seconds are taken between the two. For intelligent connected vehicles At any moment The longitudinal acceleration reflects the smoothness of vehicle acceleration and deceleration. For intelligent connected vehicles At any moment The angular velocity reflects the smoothness of a vehicle's cornering. For intelligent connected vehicles At any moment position vector , For intelligent connected vehicles The expected destination location, typically the target location of the intelligent connected vehicle after it leaves the core traffic area, is obtained from high-precision maps and navigation information. These are all weighting parameters used to balance trajectory smoothness (reducing acceleration and angular velocity) and target adherence (approaching the target point). Default values ​​can be set based on experimental experience or offline optimization, for example... .

[0090] In addition, to ensure the feasibility of the predicted driving trajectory after optimization by the trajectory optimization model, it is necessary to set some constraints on the trajectory optimization model.

[0091] (1) Vehicle dynamics constraints of a single vehicle

[0092] Manually driven vehicles cannot be remotely controlled and therefore cannot be constrained, while intelligent connected vehicles follow their vehicle dynamics models to ensure the feasibility of their trajectories.

[0093]

[0094]

[0095]

[0096] Limitations on vehicle speed, acceleration, and angular velocity:

[0097]

[0098]

[0099]

[0100] in, The minimum / maximum speed limits for intelligent connected vehicles are typically determined by traffic regulations and vehicle performance. The minimum / maximum acceleration limits for intelligent connected vehicles are typically determined by vehicle performance. The minimum / maximum angular velocity limits for intelligent connected vehicles are usually determined by vehicle performance.

[0101] (2) No collision constraints between vehicles

[0102] 1) Intelligent connected vehicles and intelligent connected vehicles

[0103] Ensure that no two intelligent connected vehicles within the cluster collide during the entire planning time domain.

[0104]

[0105] In the formula: , For intelligent connected vehicles and intelligent connected vehicles At any moment The position vector, , For intelligent connected vehicles and intelligent connected vehicles The safety radius is typically half the width and length of the vehicle, with an additional safety margin added using an expansion coefficient. The default value is 0.8 times the vehicle's geometric dimensions, plus an extra 0.5-meter safety distance. Tf is the set of all intelligent connected vehicles in the vehicle cluster, and Tf is the duration of the trajectory.

[0106] 2) Intelligent connected vehicles and manually driven vehicles

[0107] This is where the impact of human-driven vehicles is quantified and strictly enforced to ensure that intelligent connected vehicles do not collide with the predicted trajectories of human-driven vehicles.

[0108] This invention considers all high-confidence trajectory modes of manually driven vehicles during the planning process.

[0109]

[0110] In the formula: For intelligent connected vehicles At any moment The position vector, For manually driven vehicles The The predicted modal trajectory at time... The position vector, For intelligent connected vehicles and manually driven vehicles The safety radius is calculated using the same method as above. For manually driven vehicles The set of all predicted driving trajectories For manually driven vehicles The The confidence level of a predicted modal trajectory This is the confidence threshold. Only predicted trajectories of manually driven vehicles that are higher than this threshold will be included in the planning constraints to filter out low-probability trajectories. The default value is 0.6.

[0111] Since the above-mentioned collision-free constraints (especially the squared distance term) are non-convex and difficult to solve directly, this invention uses a first-order Taylor expansion to linearize them, transforming them into a series of convex constraints.

[0112] Taking intelligent connected vehicles - intelligent connected vehicle collision constraints as an example: in the first... In this iteration, for a given reference trajectory point The linearized constraints are:

[0113]

[0114] This constraint approximates the nonlinear problem as a linear inequality, and a similar linearization is applied to the collision constraint between intelligent connected vehicles and manually driven vehicles.

[0115] The linearized problem is then transformed into a quadratic programming (QP) problem, with the goal of minimizing the trajectory optimization model. Efficient solvers based on sequential convex programming algorithms (such as OSQP, Gurobi, etc.) are used for fast solving. Through multiple iterations, the predicted driving trajectory of each vehicle (including intelligent connected vehicles and manually driven vehicles) gradually approaches the optimal solution until convergence. This approach, which uses a trajectory optimization model to decompose a large-scale collaborative problem into a dynamic cluster and employs a lightweight iterative optimization method, ensures that the system can make decisions within milliseconds, meeting real-time requirements.

[0116] In addition, since the trajectory optimization model reflects the quality of a predicted driving trajectory, when the trajectory optimization model has a solution of 0, it means that the vehicle will not conflict with any other vehicle along the optimized predicted driving trajectory. Conversely, when it is not 0, it means that it will definitely conflict with other vehicles. In actual traffic scenarios, there will be two vehicles that will conflict in most cases. At this time, all two vehicles that will conflict are extracted to generate multiple conflict vehicle pairs.

[0117] Calculate the conflict risk index for each conflicting vehicle pair, set the passage priority for each vehicle in the conflicting vehicle pair based on the conflict risk index, the type of vehicle in the conflicting vehicle pair and traffic rules, and then combine the vehicles of all conflicting vehicle pairs into a passage sequence.

[0118] When a completely collision-free solution cannot be found based on the trajectory optimization model, the potential conflicts between vehicle pairs (including between intelligent connected vehicles and between intelligent connected vehicles and manually driven vehicles) are assessed to provide a basis for right-of-way decisions. By quantifying conflict risks, clear yield instructions are provided for subsequent optimization.

[0119] First, a potential conflict assessment is conducted for each vehicle pair involved in the conflict, and its conflict risk index is calculated. This metric takes into account both the time to collision (TTC) and the minimum distance (Distance).

[0120]

[0121] In the formula This is a conflict risk indicator; a higher value indicates a higher conflict risk. The typical value range is [range missing]. , The minimum collision time predicted on the predicted driving trajectory, i.e., the time required for the two vehicles to collide at the closest point, is calculated by dividing the relative distance between the two vehicles by their relative speed and taking the minimum value. The minimum distance predicted on the predicted driving trajectory, i.e., the closest distance between the two vehicles throughout the entire planning time domain, This is a weighting parameter used to balance the effects of TTC and distance. TTC is generally considered to more directly reflect dynamic risk, therefore... Assign higher weight, such as general Its value can be determined based on historical accident data or expert experience.

[0122] Then, based on the magnitude of the conflict risk index, all conflict vehicle pairs are sorted in descending order to obtain the sequence of conflicts to be processed. At the same time, create an empty yield priority list. .

[0123] Starting with the first conflicting vehicle pair in the pending conflict sequence, iterate through all conflicting vehicle pairs in the pending conflict sequence. For any conflicting vehicle pair... Since the conflicting vehicles could be two intelligent connected vehicles, or an intelligent connected vehicle and a manually driven vehicle, it is necessary to discuss each case separately.

[0124] (1) If the conflicting vehicles are If there is a conflict between intelligent connected vehicles and manually driven vehicles, then the intelligent connected vehicles... (or The vehicle was deemed to need to yield (because manually driven vehicles are uncontrollable), and was directly... (or Add to the yield priority list In, and from Remove all related to (or Related vehicle conflicts.

[0125] (2) If If the conflict involves intelligent connected vehicles, further adjudication is required. The adjudication criteria mainly include the following two aspects:

[0126] 1) Preset penalty rules: Based on preset traffic rules (e.g., straight-going vehicles have priority over left-turning vehicles, and right-turning vehicles have priority over left-turning vehicles), intelligent connected vehicles that are ordered to yield are added to the yield priority list. and from Remove all related to Related vehicle conflicts.

[0127] 2) Prioritization of Manually Driven Vehicles: In mixed traffic flows, the behavior of manually driven vehicles has higher uncertainty. Therefore, if all vehicle conflict pairs in the conflict sequence to be processed are related to intelligent connected vehicles... The number of related manually driven vehicles (i.e., possibly related to) The number of manually driven vehicles involved in conflicts, or the sum of their conflict risks, is greater than that of intelligent connected vehicles. Related manually driven vehicles, namely intelligent connected vehicles In certain scenarios, intelligent connected vehicles may have higher priority. You need to give way and allow intelligent connected vehicles to pass. Adding a vehicle to the yield priority list and vice versa, this strategy encourages connected vehicles to proactively avoid interfering with human-driven vehicles, and then... Remove all related to (or Related vehicle conflicts.

[0128] After traversing all conflicting vehicle pairs in the conflict sequence to be processed, the final yield priority list can be obtained. Based on the preset penalty rules and the second predicted driving trajectory of the manually driven vehicle, the first predicted driving trajectory of the intelligent connected vehicle is adjusted, thereby making the collaborative decision-making of the intelligent connected vehicle more robust and safe, and improving the decision-making accuracy.

[0129] The trajectory optimization model is reconstructed based on the vehicle type and traffic rules of the conflicting vehicles. Then, the reconstructed trajectory optimization model is used to optimize the predicted driving trajectory of all vehicles in the vehicle cluster, so as to obtain the optimized trajectory for each vehicle. The optimized trajectory is then sent to the corresponding vehicle.

[0130] Based on the yield priority list, the trajectory optimization model is reconstructed, and the above abstract yield instructions are transformed into specific trajectory penalty terms. These terms are then fed back into the planning stage for final optimization, ensuring that the generated trajectory can be accurately executed by the vehicle under safe conditions, thus forming a closed-loop control system.

[0131] The reconstructed trajectory optimization model is as follows:

[0132]

[0133] in, The trajectory optimization model before reconstruction As a penalty weight, For the yield priority list, For intelligent connected vehicles At any moment speed, The target speed for yielding to intelligent connected vehicle i.

[0134] Using the reconstructed trajectory optimization model, the predicted driving trajectories of all vehicles are re-optimized under constraints to obtain the optimized trajectory for each vehicle. The optimized trajectory is then sent to the corresponding intelligent connected vehicle using the RCCU. After receiving the optimized trajectory, the intelligent connected vehicle performs actions such as yielding or continuing to drive based on the optimized trajectory.

[0135] On the other hand, embodiments of the present invention also provide a signalless coordinated scheduling system for mixed traffic flows, which uses the above-mentioned signalless coordinated scheduling method for mixed traffic flows, including:

[0136] A perception module, which is used to acquire traffic perception data and send optimized trajectories to the corresponding vehicles;

[0137] The trajectory generation module is used to generate predicted driving trajectories for all vehicles based on traffic perception data, and then determine vehicles that may conflict based on the predicted driving trajectories to obtain a vehicle cluster.

[0138] A model building module, which is used to build and reconstruct trajectory optimization models;

[0139] A condition preset module, which is used to set constraints and penalty rules;

[0140] The trajectory optimization module is used to optimize the predicted driving trajectory of all vehicles in the vehicle cluster using the trajectory optimization model, extract multiple conflicting vehicle pairs that are bound to collide, set passage priorities for the vehicles in each conflicting vehicle pair, combine the vehicles of all conflicting vehicle pairs into a passage sequence, and use the reconstructed trajectory optimization model to optimize the predicted driving trajectory of all vehicles in the vehicle cluster to obtain the optimized trajectory for each vehicle.

[0141] On the other hand, embodiments of the present invention also provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the above-described method for unsignaled coordinated scheduling of mixed traffic flows.

[0142] On the other hand, embodiments of the present invention also provide a computer storage medium storing a computer program, which, when executed by a processor, causes the processor to execute the aforementioned signalless coordinated scheduling method for mixed traffic flow.

[0143] 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 signalless cooperative scheduling method for mixed traffic flow, characterized in that: include: Traffic perception data is acquired, and predicted driving trajectories of all vehicles are generated based on the traffic perception data. Then, vehicles that may conflict are identified based on the predicted driving trajectories to obtain a vehicle cluster, which includes intelligent networked vehicles and manually driven vehicles. A trajectory optimization model and constraints are constructed. Based on the constraints, the predicted driving trajectories of all intelligent connected vehicles in the vehicle cluster are optimized using the trajectory optimization model with the objective of minimizing the total cost function of the trajectory optimization model. Multiple conflicting vehicle pairs that are bound to collide are extracted. The constraints include vehicle dynamics constraints and collision-free constraints. The trajectory optimization model is as follows: in, Let the total cost function be... It is a set consisting of all intelligent connected vehicles in the vehicle cluster. To plan the time domain, For intelligent connected vehicles i at all times longitudinal acceleration, For intelligent connected vehicles i in angular velocity at time t, For intelligent connected vehicles i at all times position vector , For the desired destination position of intelligent connected vehicle i, and All are weighted parameters; Calculate the conflict risk index for each conflicting vehicle pair, and set the passage priority for each vehicle in the conflicting vehicle pair based on the conflict risk index, the type of vehicles in the conflicting vehicle pair, and traffic rules. A yield priority list is constructed based on the passage priority of each vehicle, and the trajectory optimization model is reconstructed based on the yield priority list. The vehicles in the priority list are those determined to yield. The reconstructed trajectory optimization model is as follows: in, The total cost function of the trajectory optimization model before reconstruction. As a penalty weight, For the yield priority list, For intelligent connected vehicles At any moment speed, The target speed for yielding to intelligent connected vehicle i; Then, the reconstructed trajectory optimization model is used to optimize the predicted driving trajectory of all intelligent connected vehicles in the vehicle cluster, obtaining the optimized trajectory for each vehicle, and then the optimized trajectory is sent to the corresponding vehicle.

2. The signalless coordinated scheduling method for mixed traffic flow as described in claim 1, characterized in that: Before generating the predicted driving trajectory, the speed of all intelligent connected vehicles is adjusted based on traffic perception data to ensure that they drive at the optimal speed. For intelligent connected vehicles, a position change function is constructed based on traffic perception data, and a first predicted driving trajectory is generated based on the position change function. For manually driven vehicles, multiple second predicted driving trajectories and corresponding confidence scores are generated based on the GAT learning model using traffic perception data. Then, all second predicted driving trajectories with confidence scores less than the confidence score threshold are eliminated.

3. The signalless coordinated scheduling method for mixed traffic flow as described in claim 2, characterized in that: A method for identifying potential vehicle clusters based on predicted driving trajectories: The intersection points between all trajectories are determined based on the predicted driving trajectories of all vehicles, and multiple conflict zones are generated with the intersection points as the center. Calculate the predicted conflict time when all intelligent connected vehicles reach any conflict area while traveling along their corresponding predicted trajectories. Create a vehicle cluster. For any two intelligent connected vehicles, if the minimum distance between their corresponding predicted trajectories is less than the safe distance and the difference in conflict time is less than the safe time, then add the two intelligent connected vehicles to the vehicle cluster. For any intelligent connected vehicle and any manually driven vehicle, if the first predicted driving trajectory of the intelligent connected vehicle and any second predicted driving trajectory of the manually driven vehicle intersect in the same conflict area, then the intelligent connected vehicle and the manually driven vehicle are added to the vehicle cluster.

4. The signalless coordinated scheduling method for mixed traffic flow as described in claim 2, characterized in that: The collision-free constraint between two intelligent connected vehicles is: in, Intelligent connected vehicles and intelligent connected vehicles At any moment The position vector, , For intelligent connected vehicles and intelligent connected vehicles The safety radius, The set consists of all intelligent connected vehicles in the vehicle cluster, and Tf is the trajectory duration; The collision-free constraint between intelligent connected vehicles and manually driven vehicles is: In the formula: For intelligent connected vehicles At any moment The position vector, For manually driven vehicles The The predicted driving trajectory at time... The position vector, For intelligent connected vehicles and manually driven vehicles The safety radius, For manually driven vehicles The set of all predicted driving trajectories For manually driven vehicles The The confidence level of the predicted driving trajectory. This is the confidence threshold.

5. The signalless coordinated scheduling method for mixed traffic flow as described in claim 2, characterized in that: Methods for setting traffic priorities for each conflicting vehicle pair based on conflict risk indicators, the types of vehicles in the conflicting vehicle pair, and traffic rules include: Based on the magnitude of the conflict risk index, all conflict vehicle pairs are sorted in descending order to obtain the conflict sequence to be processed. Create a yield priority list, iterate through all the first conflict vehicle pairs in the pending conflict sequence, and in response to the conflict vehicle pair being intelligent connected vehicle i and human-driven vehicle m, add intelligent connected vehicle i to the yield priority list, and then remove all conflict vehicle pairs related to intelligent connected vehicle i from the pending conflict sequence. In response to a conflicting vehicle pair being intelligent connected vehicle i and intelligent connected vehicle j, intelligent connected vehicle i or intelligent connected vehicle j is added to the yield priority list according to preset traffic rules, and then all conflicting vehicle pairs related to intelligent connected vehicle i or intelligent connected vehicle j are removed from the pending conflict sequence.

6. The signalless coordinated scheduling method for mixed traffic flow as described in claim 5, characterized in that: When both conflicting vehicle pairs are intelligent connected vehicles and intelligent connected vehicle i needs to be added to the yield priority list according to preset traffic rules, if the number of manually driven vehicles associated with intelligent connected vehicle i in the conflict sequence to be processed is greater than the number of manually driven vehicles associated with intelligent connected vehicle j, then intelligent connected vehicle j is added to the yield priority list, and then all conflicting vehicle pairs associated with intelligent connected vehicle j are deleted from the conflict sequence to be processed.

7. A signalless cooperative scheduling system for mixed traffic flow, characterized in that: The system uses a signalless cooperative scheduling method for mixed traffic flow as described in any one of claims 1-6, comprising: A perception module, which is used to acquire traffic perception data and send optimized trajectories to the corresponding vehicles; The trajectory generation module is used to generate predicted driving trajectories for all vehicles based on traffic perception data, and then determine vehicles that may conflict based on the predicted driving trajectories to obtain a vehicle cluster. A model building module, which is used to build and reconstruct trajectory optimization models; A condition preset module, which is used to set constraints and penalty rules; The trajectory optimization module is used to optimize the predicted driving trajectory of all intelligent connected vehicles in the vehicle cluster using the trajectory optimization model, extract multiple conflicting vehicle pairs that are bound to conflict, set the passage priority for the vehicles in each conflicting vehicle pair, then combine the vehicles of all conflicting vehicle pairs into a passage sequence, and use the reconstructed trajectory optimization model to optimize the predicted driving trajectory of all intelligent connected vehicles in the vehicle cluster to obtain the optimized trajectory for each vehicle.

8. A computer device, characterized in that: It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform a signalless coordinated scheduling method for mixed traffic flow as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that: The system contains a computer program that, when executed by a processor, causes the processor to perform a signalless coordinated scheduling method for mixed traffic flows as described in any one of claims 1 to 6.

Images