A vehicle dispatching method for optimizing traffic load on main roads at intersections

Abstract

The present invention provides a vehicle dispatching method for optimizing traffic load on main roads at intersections, including: a cloud network constructs a global optimization model and models of each main road; a roadside unit constructs and sends a first traffic load optimization model; an intelligent vehicle constructs and uploads its own optimization model based on the global optimization model, the models of each main road and the first traffic load optimization model; the roadside unit receives the vehicle's own optimization model, integrates it into a second traffic load optimization model and sends it; an edge computing unit receives the second traffic load optimization model, solves the optimization result, establishes a target main road congestion prediction model, sends the optimization result and the target main road congestion prediction model; the cloud network receives the optimization result and gives an intelligent vehicle dispatching decision. The present invention solves the technical problem that there are security risks in the unencrypted data transmission between the intelligent transportation system and the vehicle, and achieves the technical effect of effectively protecting the local privacy data of the intelligent vehicle.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a vehicle dispatching method for optimizing traffic load on main roads at intersections based on a vehicle networking platform. Background Art

[0002] Vehicle-to-everything (V2X) technology is a new generation of information and communication technology that connects vehicles to everything. Through the communication and interaction between vehicles, vehicles and people, vehicles and roadside infrastructure, and vehicles and cloud networks, it realizes real-time and dynamic interaction of all-time and space traffic information, meets the requirements of active vehicle safety control and road collaborative management, and achieves effective collaboration among people, vehicles, and roads. With the support of the Internet of Vehicles, intelligent vehicles and cloud networks, including roadside units, have achieved rapid development. Intelligent vehicles can achieve multi-source heterogeneous data fusion and state prediction through mutual collaboration with roadside units and cloud networks by using various sensors, on-board hardware and software equipment, etc., to meet the requirements of vehicle environmental perception enhancement and expansion, reduce the accident rate, and achieve safe and efficient driving. Intelligent vehicles, roadside units, and cloud networks can achieve joint decision-making and control through mutual collaboration, avoid queuing, waiting, and decision logic deadlock in the traffic process, improve traffic efficiency, and form a safe, efficient, and environmentally friendly intelligent transportation system.

[0003] However, in actual applications, there are several problems: generally, the communication data between the cloud network and the smart vehicle is the original vehicle data connected through wireless communication, including the destination of the smart vehicle, user binding information, vehicle driving status, etc., which may be intercepted by malicious users during the transmission process, and illegal acts such as deletion and modification of vehicle data may be committed, or serious privacy leakage may occur. Therefore, there are security risks in the unencrypted data transmission between the intelligent transportation system and the vehicle; generally, traffic congestion is alleviated by adaptively adjusting the phase time of traffic lights. When there are multiple traffic congestion situations on main roads that need to be optimized, simply adaptively adjusting the phase time of traffic lights to alleviate traffic congestion is too inefficient and difficult to effectively reduce the traffic load on the main roads; in the prior art, the available computing resources of the cloud network and smart vehicles are limited. Too many tasks on the cloud network will seriously affect the actual system deployment of the Internet of Vehicles, and computing and processing problems from the cloud and then sending them to each smart car will cause large delays, further reducing the smoothness of traffic on the main roads. Summary of the invention

[0004] Therefore, the present invention provides a vehicle scheduling method for optimizing traffic load on main roads at intersections, which solves the technical problem of potential security risks in unencrypted data transmission between cloud networks and vehicles.

[0005] In order to solve the above problems, the present invention provides a vehicle dispatching method for optimizing the traffic load of the main road at the intersection, comprising:

[0006] Step S1: The cloud network publishes the target trunk optimization task and builds the global optimization model M ITS And the main road models

[0007] Step S2: The roadside unit constructs the first traffic load optimization model Send the first traffic load optimization model to smart vehicles;

[0008] Step S3: The intelligent vehicle receives the target trunk road optimization task and calculates the target trunk road optimization task according to the global optimization model M. ITS , main road models and the first traffic load optimization model Constructing the vehicle's own optimization model Upload the vehicle's own optimized model To the roadside unit;

[0009] Step S4: The roadside unit receives the vehicle's own optimization model According to the first traffic load optimization model Integrate into the second traffic load optimization model Send the second traffic load optimization model to the edge computing unit;

[0010] Step S5: The edge computing unit receives the second traffic load optimization model Solve the optimization results and establish the target trunk road congestion prediction model Send optimization results and target trunk road congestion prediction model to cloud networks, roadside units and smart road signs;

[0011] Step S6: The cloud network receives the optimization results and aggregates them into a global optimization result, compares the global optimization result with the target threshold, and if the global optimization result does not reach the target threshold, repeats steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision;

[0012] Among them, the second traffic load optimization model The first traffic load optimization model with specific values ​​after integration

[0013] Compared with the existing technology, the technical effect achieved by adopting this technical solution is as follows: the cloud network publishes the target trunk road optimization task to wake up the roadside unit, intelligent vehicle and edge computing unit, and builds a global optimization model M on the cloud network side. ITSAnd the main road models Provide an initial model for the roadside unit and edge computing unit, and the roadside unit and edge computing unit can build other models based on it; the roadside unit uses the global optimization model M sent by the cloud network ITS And the main road models Building a traffic load optimization model Traffic load optimization model Sent to the smart vehicle to provide the required parameters; the smart vehicle receives the traffic load optimization model transmitted by the roadside unit Optimize the model based on traffic load The required parameters in the calculation are used to calculate the vehicle model and build the vehicle's own optimization model Upload the vehicle's own optimized model The roadside unit transmits the vehicle's own optimization model transmitted by the intelligent vehicle Optimize the model based on traffic load Integrate into the second traffic load optimization model The edge computing unit receives the second traffic load optimization model Solve the optimization results and build a target trunk road congestion prediction model The cloud network receives the optimization results and aggregates them into global optimization results. The global optimization results are compared with the target threshold. If the global optimization results do not reach the target threshold, steps S2-S5 are repeated. If the global optimization results reach the target threshold, intelligent vehicle scheduling decisions are given, including left turn, right turn and straight driving. By uploading the vehicle's own optimization model To replace information such as vehicle destination, it can effectively protect the local privacy data of smart vehicles.

[0014] The system of the present invention includes a cloud network and all roadside units and edge computing units within a region, and the roadside units are responsible for communicating with smart vehicles with wireless communication. Because the specific model solution calculation is delegated to the edge computing unit and the roadside unit, the cloud network can expand the regional coverage, receive the roadside units and edge computing units in the new area, and form a new cloud network. When the data throughput of the cloud network system is sufficient, the traffic load of the main roads in the entire region or city can be optimized. The overall cloud network system has a certain scalability and can be expanded from one region to the optimization of the traffic load of the entire region.

[0015] Furthermore, in the above technical solution, the cloud network publishes the target trunk road optimization task and constructs the global optimization model M ITS And the main road models include:

[0016] Step S11: The cloud network constructs a global optimization model MITS And the main road models And initialize the global optimization model M ITS And the main road models

[0017] Building a global optimization model Constructing models of main roads

[0018] Among them, R i is the number of roadside unit mappings corresponding to the main road i; P is the number of main road sets; i is the number of a target main road, the total number is I; j is the number of a roadside unit, the total number is J;

[0019] Step S12: The cloud network publishes the target trunk road optimization task to the roadside unit, edge computing unit and smart vehicle;

[0020] Step S13: Send the global optimization model M ITS And the main road models To the roadside unit and edge computing unit, and assign the corresponding main roads to the roadside unit and edge computing unit.

[0021] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: the cloud network first initializes the trunk road set P in the area = {P1, P2, ..., P i},i max is the maximum number of main roads in the region, i is the number of a target main road; I is the total number of target main roads in the region; P1 represents the first main road in the region, P i represents the i-th main road in the area; and establishes a mapping between roadside units and main roads R = {R1, R2, ..., R j}, R is the trunk line mapping set, j is the roadside unit number; J is the total number of roadside units, R1 represents the first roadside unit on a trunk road, R j represents the jth roadside unit on the main road, Represents the target trunk road optimization model collected by the jth roadside unit; the cloud network builds the model of each trunk road represents the optimization model of the i-th target trunk road. Assuming that there are i trunk roads in an area and j roadside units on a trunk road, the trunk road models represents the target trunk road optimization model set collected by all roadside units from the first roadside unit to the jth roadside unit on the i-th trunk road; the cloud network constructs a global optimization model M ITS , the global optimization model M ITSRepresents the set of trunk optimization models collected from all trunks in the region, from the first trunk to the i-th trunk. The cloud network publishes the target trunk optimization task to the roadside unit, edge computing unit and intelligent vehicle. The roadside unit, edge computing unit and intelligent vehicle receive the target trunk optimization task and start working according to the demand. The cloud network sends the global optimization model M ITS And the main road models To the roadside unit and edge computing unit, the roadside unit receives the global optimization model M ITS And the main road models and transmitted to the intelligent vehicle; and according to the global optimization model M ITS And the main road models Assigning corresponding main roads to roadside units and edge computing units, and delegating the solution of the optimization task model to edge computing units and intelligent vehicles can make full use of the existing computing resources in the road environment, solve the traffic load optimization of each main road in the entire area, and improve the efficiency of computing resource utilization.

[0022] Furthermore, in the above technical solution, the global optimization model M is sent ITS And the main road models To the roadside unit and edge computing unit, and allocate the corresponding main roads to the roadside unit and edge computing unit, including:

[0023] Step S130: The cloud network sends the global optimization model M ITS And the main road models To roadside units and edge computing units;

[0024] Step S131: Initialize the number of cluster centers and the locations of cluster centers;

[0025] Step S132: Calculate the similarity (transmission distance) between each main road and the cluster center, and classify the main roads into the most similar class;

[0026] Step S133: updating the cluster center;

[0027] Step S134: If the cluster centers have not converged, repeat steps S132-S133;

[0028] Step S135: establishing an optimization demand matrix for each cluster;

[0029] Step S136: Initialize the supply matrix of the edge computing unit for the optimization task;

[0030] Step S137: replanning the demand matrix of each cluster, and adjusting the clustering of some trunk road optimization tasks;

[0031] Step S138: If the supply of edge computing units does not satisfy the demand matrix, repeat step S137;

[0032] Step S139: Establish an edge computing selection model and allocate edge computing units to each cluster group.

[0033] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: clustering algorithm is used as the task allocation framework of edge computing unit, the number of cluster centers K is initialized, representing the number of target trunk road sets, and the farthest K trunk roads in the area are set as the starting cluster center position C; the distance D between each trunk road and the cluster center is calculated, and each trunk road is classified into the cluster group with the closest distance; the average position of each cluster group is calculated, and the cluster center C* is updated; if the cluster center does not converge, that is, the cluster center will still be updated as needed, and the steps are repeated to calculate the distance between each trunk road and the cluster center. Distance D, classify each main road into the cluster grouping with the nearest distance and calculate the average position of each cluster grouping, update the cluster center C*; establish the optimization demand matrix of each cluster; initialize the supply matrix of the edge computing unit for the optimization task; re-plan the demand matrix of each cluster, and adjust the cluster grouping of some main road optimization tasks; if the supply of the edge computing unit does not meet the demand matrix, repeat the steps to re-plan the demand matrix of each cluster, adjust the cluster grouping of some main road optimization tasks; establish the edge computing unit selection model, and allocate cluster groups according to the principle of the shortest transmission distance.

[0034] Furthermore, in the above technical solution, the roadside unit constructs a first traffic load optimization model Send the first traffic load optimization model To smart vehicles, including:

[0035] Step S21: The roadside unit receives the global optimization model M ITS And the main road models The global optimization model M ITS And the main road models Delivered to smart vehicles;

[0036] Step S22: The roadside unit calculates the global optimization model M ITS And the main road models Constructing the first traffic load optimization model in, Optimizing models for the vehicle itself; To map RSU j The number of intelligent vehicles; u is the number of intelligent vehicles;

[0037] Step S23: The roadside unit sends the first traffic load optimization model To smart vehicles.

[0038] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: the roadside unit is responsible for communicating with the intelligent vehicle with wireless communication, and the roadside unit receives the global optimization model M transmitted from the cloud network. ITS And the main road models And the global optimization model M ITS And the main road models Transferring the model solution to the smart vehicle can make full use of the existing computing resources of the smart vehicle, relieve the computing pressure of the cloud network, and improve the efficiency of computing resource utilization; represents the vehicle optimization model of the u-th vehicle, To map RSU j The number of intelligent vehicles, u is the number of intelligent vehicles, and the roadside unit constructs the first traffic load optimization model The first traffic load optimization model Represents the vehicle optimization model collected by the first intelligent vehicle to the uth intelligent vehicle on the jth roadside unit; the roadside unit sends the traffic load optimization model To smart vehicles, provide the initial first traffic load optimization model for smart vehicles Intelligent vehicles can optimize the model according to the first traffic load The required parameters in are used to calculate the ego-vehicle model.

[0039] Furthermore, in the above technical solution, the intelligent vehicle receives the target trunk road optimization task and optimizes the target trunk road according to the global optimization model M ITS , main road models and the first traffic load optimization model Constructing the vehicle's own optimization model Upload the vehicle's own optimized model To the roadside unit, including:

[0040] Step S31: The intelligent vehicle receives the target trunk road optimization task issued by the cloud network;

[0041] Step S32: The intelligent vehicle receives the global optimization model M transmitted by the roadside unit ITS And the main road models

[0042] Step S33: The intelligent vehicle receives the first traffic load optimization model constructed by the roadside unit

[0043] Step S34: the intelligent vehicle initializes the local driving state of the intelligent vehicle and collects external environment information;

[0044] Step S35: the intelligent vehicle replans the destination navigation route;

[0045] Step S36: Establishing an intersection selection priority list and corresponding cost model (c turn ,c path ,c traffic );

[0046] Step S37: Construct the priority list and the corresponding cost model into the vehicle's own optimization model Upload to the roadside unit;

[0047] Among them, the intersection options include: left turn, straight ahead, right turn, cost model (c turn ,c path ,c traffic ) includes turning cost, distance cost, and predicted traffic cost of the selected road. (c turn ) is the cost model of the left turn cost, the distance cost, and the predicted traffic volume cost of the selected road, (c path ) is the cost model of the straight turning cost, the distance cost, and the predicted traffic volume cost of the selected road, (c traffic ) is a cost model that includes the turning cost of turning right, the distance cost, and the predicted traffic volume cost of the selected road.

[0048] Compared with the existing technology, the technical effect achieved by adopting this technical solution is as follows: it is mainly composed of intelligent vehicles that can be connected to the Internet of Vehicles. The intelligent vehicles receive the target trunk road optimization tasks issued by the cloud network, start preparation work after receiving the optimization tasks, and receive the global optimization model M transmitted by the roadside unit. ITS , main road models The first traffic load optimization model constructed with roadside units In order to calculate the ego vehicle model according to the required parameters in the model; the intelligent vehicle initializes the ego vehicle state model, and continuously collects the information required for route planning near the road according to the various types of sensors and other equipment installed on it, including various types of single / binocular cameras, ultrasonic radars, lidars, and vehicle driving status (speed, acceleration, reference trajectory, etc.), and establishes the corresponding environment model; then establishes the priority list and loss cost model for the three road options of left turn, straight, and right turn (c turn ,c path ,c traffic), the cost model includes turning cost, distance cost, and predicted traffic cost of the selected road; the route is replanned based on the driving route navigation algorithm used by the vehicle itself, and the comprehensive optimal driving route under the three intersection road options of left turn, straight and right turn is calculated respectively. The optimal route aims to minimize the loss cost. According to the loss cost of the three road options, a priority list with the lowest cost and the corresponding cost model are established; the road selection priority list and the corresponding cost model are constructed as the vehicle's own optimization model The re-route is calculated locally in the smart vehicle, and the intersection selection priority and vehicle optimization model are uploaded. To replace the vehicle destination information, effectively protect the local privacy data of smart vehicles. And during the transmission process, for the same smart vehicle, for each main road it passes through, the intersection selection priority and vehicle optimization model received by the roadside unit It has no conversion relationship and is only related to the current main road intersection, which further strengthens the ability to prevent malicious attackers from cracking user privacy data.

[0049] Furthermore, in the above technical solution, the roadside unit receives the vehicle's own optimization model According to the first traffic load optimization model Integrate into the second traffic load optimization model Send the second traffic load optimization model To the edge computing unit, including:

[0050] Step S41: The roadside unit receives the vehicle self-optimization model established by the intelligent vehicle

[0051] Step S42: The roadside unit collects the driving status of non-intelligent vehicles

[0052] Step S43: The roadside unit optimizes the model based on the vehicle itself and non-intelligent vehicle driving status Integrate into the second traffic load optimization model

[0053] Step S44: The roadside unit sends the second traffic load optimization model To the edge computing unit.

[0054] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: the roadside unit receives the vehicle self-optimization model established by the intelligent vehicle Roadside units collect driving status of non-intelligent vehicles Including lane, speed, and vehicle optimization model and non-intelligent vehicle driving status Integrate into the second traffic load optimization model The second traffic load optimization model To the edge computing unit for accumulation and aggregation.

[0055] Furthermore, in the above technical solution, the edge computing unit receives the second traffic load optimization model Solve the optimization results and establish the target trunk road congestion prediction model Send optimization results and target trunk road congestion prediction model To cloud networks, roadside units and smart road signs, including:

[0056] Step S51: The edge computing unit receives the global optimization model M sent by the cloud network ITS And the main road models

[0057] Step S52: The edge computing unit receives the second traffic load optimization model sent by the roadside unit And aggregate and accumulate to generate the second main road model

[0058] Step S53: The edge computing unit optimizes the global model M ITS , Vehicle self-optimization model And the second main road model Iteratively solve the optimization results;

[0059] Step S54: Establishing a target trunk road congestion prediction model based on the optimization results

[0060] Step S55: The edge computing unit sends the optimization results and the target trunk road congestion prediction model to cloud networks, roadside units and smart road signs.

[0061] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: the edge computing unit receives the initialized global optimization model M sent by the cloud network ITS And initialize the main road model The edge computing unit receives the second traffic load optimization model sent by the roadside unit The second traffic load optimization model Represents the set of vehicle optimization models collected by the jth roadside unit in the area; and aggregates and accumulates to generate the second main road model The second main road model Represents the set of traffic load optimization models collected by all roadside units on the i-th trunk road in the area; iterates to solve the optimization result, which includes one of the results of left turn, right turn, and straight driving. The edge computing unit establishes the target trunk road congestion prediction model based on the optimization result Including the target trunk road throughput within the cycle; finally, the optimization results and the target trunk road congestion prediction model are combined It is sent to the cloud network, roadside units and smart road signs, only considering the traffic load optimization of each main road, and assigning the optimization solution task to the edge computing unit, which can reduce the computing pressure of the cloud network and improve the solution space and speed.

[0062] Furthermore, in the above technical solution, the edge computing unit is based on the global optimization model M ITS , Vehicle self-optimization model And the second main road model Iterative solution optimization results, including:

[0063] Step S531: Initialize the global optimization model M ITS and load optimization models;

[0064] Step S532: constructing an intelligent vehicle fork junction selection model;

[0065] Step S533: Select the intelligent vehicle fork with the highest priority as the initial solution;

[0066] Step S534: adjusting the selection priority list of the intelligent vehicle according to the fork capacity;

[0067] Step S535: Iterate step S534 until the algorithm converges to the minimum cost that meets the traffic capacity of the fork intersection and obtains the optimization result.

[0068] Compared with the prior art, the technical effects achieved by adopting the technical solution are as follows: the present invention adopts a delayed receiving algorithm to dispatch intelligent vehicles with lower cost losses to update road selection, initialize the traffic capacity model and load optimization model of the main road fork; takes the current road selection of each intelligent vehicle as the initial solution X, updates the intelligent vehicle road selection priority list and the loss cost model, selects the vehicle with the smallest loss cost due to changing the road selection for road update, updates the global load optimization result, and verifies the maximum passable vehicle capacity of the target main road; if the global load optimization result or the maximum passable vehicle capacity of the target main road does not meet the target threshold, repeatedly updates the global load optimization result, and if the algorithm converges ahead of time, transmits the load optimization result; if the number of iterations reaches the maximum value, stops the above steps.

[0069] Furthermore, in the above technical solution, the edge computing unit sends the optimization results and the target trunk road congestion prediction model To cloud networks, roadside units and smart road signs, including:

[0070] Step S551: The edge computing unit sends the optimization results and the target trunk road congestion prediction model To cloud network and roadside units;

[0071] Step S552: The edge computing unit uses the target trunk road congestion prediction model Delivered to smart road signs;

[0072] Step S553: ​​Displaying the road congestion situation on the intelligent road sign.

[0073] Compared with the existing technology, the technical effect achieved by adopting this technical solution is: the edge computing unit sends the optimization results and the target trunk road congestion prediction model The optimization results are sent to the edge computing unit through the cloud network. The edge computing unit transmits the optimization results to the smart vehicle through the roadside unit, alleviating the delay problem of the vehicle receiving the optimization results from the cloud network. The smart vehicle chooses whether to make a scheduling decision based on the global optimization results aggregated by the cloud network and the judgment results of the target threshold. The edge computing unit transmits the target trunk road congestion prediction model to the smart vehicle. The information is transmitted to the intelligent road signs, and the road congestion situation is displayed on the intelligent road signs. Thus, the road congestion situation provides a suggestion for the road selection of non-intelligent vehicles. Non-intelligent vehicles can choose roads according to the road congestion situation displayed on the intelligent road signs, thereby further alleviating the congestion on the main roads.

[0074] Furthermore, in the above technical solution, the cloud network receives the optimization results and aggregates them into a global optimization result, compares the global optimization result with the target threshold, and if the global optimization result does not reach the target threshold, repeats steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision, including:

[0075] Step S61: The cloud network receives the optimization results and aggregates them into a global optimization result;

[0076] Step S62: Compare the global optimization result with the target threshold;

[0077] Step S63: Based on the comparison between the global optimization result and the target threshold, an intelligent vehicle dispatching decision is made or the instructions of steps S2-S5 are repeated. If the global optimization result does not reach the target threshold, the cloud network gives an instruction to repeat steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle dispatching decision.

[0078] Compared with the prior art, the technical effect achieved by adopting this technical solution is as follows: the cloud network receives the optimization results of the edge computing unit optimization solution, and aggregates them into a global optimization result in the cloud network. According to the comparison result between the global optimization result and the target threshold, the cloud network makes an instruction. If the global optimization result meets the target threshold, the intelligent vehicle receives the optimization result transmitted by the roadside unit and makes an intelligent vehicle scheduling decision. If the global optimization result does not meet the target threshold, the cloud network issues a repeat instruction and repeats steps S2-S5.

[0079] After adopting the technical solution of the present invention, the following technical effects can be achieved:

[0080] (1) Effectively protect the local privacy data of smart vehicles;

[0081] (2) It can make full use of the existing computing resources in the road environment, improve the efficiency of computing resource utilization, and reduce the computing pressure of the cloud network;

[0082] (3) Improve the speed at which vehicles receive optimization results;

[0083] (4) The overall intelligent transportation system has a certain degree of scalability and can be expanded from one area to the entire area for traffic load optimization;

[0084] (5) The vehicle traffic model of each main road and the road selection model of intelligent vehicles can be trained using deep reinforcement learning networks such as DQN and RNN. Through early training, the use of later computing resources can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 This is one of the flow charts of the steps of a vehicle dispatching method for optimizing traffic load on main roads at intersections according to some embodiments of the present invention.

[0086] Figure 2 This is a second flow chart of the steps of a vehicle dispatching method for optimizing traffic load on main roads at intersections according to some embodiments of the present invention.

[0087] Figure 3 This is a flowchart of the steps of a vehicle dispatching method for optimizing traffic load on main roads at intersections according to some embodiments of the present invention.

[0088] Figure 4 This is a fourth step flow chart of a vehicle dispatching method for optimizing traffic load on main roads at intersections according to some embodiments of the present invention.

[0089] Figure 5 This is a flowchart of the fifth step of a vehicle dispatching method for optimizing traffic load on main roads at intersections according to some embodiments of the present invention. DETAILED DESCRIPTION

[0090] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0091] See also Figure 1 The present invention provides a vehicle dispatching method for optimizing traffic load on a main road at an intersection, comprising:

[0092] Step S1: The cloud network publishes the target trunk optimization task and builds the global optimization model M ITS And the main road models

[0093] Step S2: The roadside unit constructs the first traffic load optimization model Send the first traffic load optimization model to smart vehicles;

[0094] Step S3: The intelligent vehicle receives the target trunk road optimization task and calculates the target trunk road optimization task according to the global optimization model M. ITS , main road models and the first traffic load optimization model Constructing the vehicle's own optimization model Upload the vehicle's own optimized model To the roadside unit;

[0095] Step S4: The roadside unit receives the vehicle's own optimization model According to the first traffic load optimization model Integrate into the second traffic load optimization model Send the second traffic load optimization model to the edge computing unit;

[0096] Step S5: The edge computing unit receives the second traffic load optimization model Solve the optimization results and establish the target trunk road congestion prediction model Send optimization results and target trunk road congestion prediction model to cloud networks, roadside units and smart road signs;

[0097] Step S6: The cloud network receives the optimization results and aggregates them into a global optimization result, compares the global optimization result with the target threshold, and if the global optimization result does not reach the target threshold, repeats steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision;

[0098] Among them, the second traffic load optimization model The first traffic load optimization model with specific values ​​after integration

[0099] The solution of this embodiment is applied to the scenario of smart vehicles avoiding congested roads during peak hours. The most common situation is that after the smart vehicle determines the destination, the driving route navigation used by the smart vehicle itself starts to plan the path. However, due to too many vehicles on the road, driving according to the driving route navigation used by the vehicle itself may encounter traffic congestion on the main road, and it is easy to cause the traffic congestion on the main road to become more serious.

[0100] In this embodiment, the cloud network needs to first publish the target trunk road optimization task and initialize the global optimization model M ITS And the main road models Constructing the global optimization model M ITS And the main road models Provide an initial model for the roadside unit and edge computing unit, on which the roadside unit and edge computing unit can build other models; use the cloud network to send tasks to the roadside unit, edge computing unit and smart vehicle, and delegate the optimization task model to the edge computing unit and smart vehicle for calculation, thereby alleviating the computing pressure of the cloud network, making full use of the existing computing resources in the road environment, reducing the waste of resources, and sending the calculation results to the smart vehicle more quickly, alleviating the delay problem, and further accelerating the traffic load optimization speed of each main road in the region.

[0101] The roadside unit uses the global optimization model M sent by the cloud network ITS And the main road models Building a traffic load optimization model Traffic load optimization model Sent to the smart vehicle to provide the required parameters; the smart vehicle receives the traffic load optimization model transmitted by the roadside unit Optimize the model based on traffic load The required parameters in the calculation are used to calculate the vehicle model and build the vehicle's own optimization model Upload the vehicle's own optimized model The roadside unit transmits the vehicle's own optimization model transmitted by the intelligent vehicle Optimize the model based on traffic load Integrate into the second traffic load optimization model The edge computing unit receives the second traffic load optimization model Solve the optimization results and build a target trunk road congestion prediction model The cloud network receives the optimization results and aggregates them into global optimization results. The global optimization results are compared with the target threshold. If the global optimization results do not reach the target threshold, steps S2-S5 are repeated. If the global optimization results reach the target threshold, an intelligent vehicle scheduling decision is given, where the intelligent vehicle scheduling decision includes left turn, right turn and straight going.

[0102] Furthermore, the cloud network publishes the target trunk optimization task and constructs the global optimization model M ITS And the main road models include:

[0103] Step S11: The cloud network constructs a global optimization model M ITS And the main road models And initialize the global optimization model M ITS And the main road models

[0104] Building a global optimization model Constructing models of main roads

[0105] Among them, R i is the number of roadside unit mappings corresponding to the main road i; P is the number of main road sets; i is the number of a target main road, the total number is I; j is the number of a roadside unit, the total number is J;

[0106] Step S12: The cloud network publishes the target trunk road optimization task to the roadside unit, edge computing unit and smart vehicle;

[0107] Step S13: Send the global optimization model M ITS And the main road models To the roadside unit and edge computing unit, and assign the corresponding main roads to the roadside unit and edge computing unit.

[0108] In this embodiment, the cloud network first initializes the trunk road set P in the area = {P1, P2, ..., P i}, i max is the maximum number of main roads in the region, i is the number of a target main road; I is the total number of target main roads in the region; P1 represents the first main road in the region, P i represents the i-th main road in the area; and establishes a mapping between roadside units and main roads R = {R1, R2, ..., R j}, R is the trunk line mapping set, j is the roadside unit number; J is the total number of roadside units, R1 represents the first roadside unit on a trunk road, R j represents the jth roadside unit on the main road, Represents the target trunk road optimization model collected by the jth roadside unit; the cloud network builds the model of each trunk road represents the optimization model of the i-th target trunk road. Assuming that there are i trunk roads in an area and j roadside units on a trunk road, the trunk road models represents the target trunk road optimization model set collected by all roadside units from the first roadside unit to the jth roadside unit on the i-th trunk road; the cloud network constructs a global optimization model M ITS , the global optimization model M ITS Represents the set of trunk optimization models collected from all trunk roads in the region, from the first trunk road to the i-th trunk road. The cloud network publishes the target trunk optimization task to the roadside unit, edge computing unit and intelligent vehicle. The roadside unit, edge computing unit and intelligent vehicle receive the target trunk optimization task and start working according to the demand. The cloud network sends the global optimization model M ITS And the main road models To the roadside unit and edge computing unit, the roadside unit receives the global optimization model M ITS And the main road models and transmitted to the intelligent vehicle; and according to the global optimization model M ITS And the main road models Assigning corresponding main roads to roadside units and edge computing units, and delegating the solution of the optimization task model to edge computing units and intelligent vehicles can make full use of the existing computing resources in the road environment and solve the traffic load optimization of each main road in the entire area.

[0109] Further, send the global optimization model M ITS And the main road models To the roadside unit and edge computing unit, and allocate the corresponding main roads to the roadside unit and edge computing unit, including:

[0110] Step S130: The cloud network sends the global optimization model M ITS And the main road models To roadside units and edge computing units;

[0111] Step S131: Initialize the number of cluster centers and the locations of cluster centers;

[0112] Step S132: Calculate the similarity (transmission distance) between each main road and the cluster center, and classify the main roads into the most similar class;

[0113] Step S133: updating the cluster center;

[0114] Step S134: If the cluster centers have not converged, repeat steps S132-S133;

[0115] Step S135: establishing an optimization demand matrix for each cluster;

[0116] Step S136: Initialize the supply matrix of the edge computing unit for the optimization task;

[0117] Step S137: replanning the demand matrix of each cluster, and adjusting the clustering of some trunk road optimization tasks;

[0118] Step S138: If the supply of edge computing units does not satisfy the demand matrix, repeat step S137;

[0119] Step S139: Establish an edge computing selection model and allocate edge computing units to each cluster group.

[0120] In this embodiment, a clustering algorithm is used as the task allocation framework of the edge computing unit, the number of cluster centers K is initialized, representing the number of target trunk road sets, and the farthest K trunk roads in the area are set as the starting cluster center position C; the distance D between each trunk road and the cluster center is calculated, and each trunk road is classified into the cluster group with the nearest distance; the average position of each cluster group is calculated, and the cluster center C* is updated; if the cluster center has not converged, that is, the cluster center will still be updated as needed, the steps of calculating the distance D between each trunk road and the cluster center, classifying each trunk road into the cluster group with the nearest distance and calculating the average position of each cluster group are repeated, and the cluster center C* is updated; an optimization demand matrix for each cluster is established; the supply matrix of the edge computing unit for the optimization task is initialized; the demand matrix of each cluster is re-planned, and the cluster grouping of some trunk road optimization tasks is adjusted; if the supply of the edge computing unit does not meet the demand matrix, the steps of re-planning the demand matrix of each cluster are repeated, and the cluster grouping of some trunk road optimization tasks is adjusted; an edge computing unit selection model is established, and cluster groups are allocated according to the principle of the shortest transmission distance.

[0121] Furthermore, the roadside unit constructs the first traffic load optimization model Send the first traffic load optimization model To smart vehicles, including:

[0122] Step S21: The roadside unit receives the global optimization model M ITS And the main road models The global optimization model M ITS And the main road models Delivered to smart vehicles;

[0123] Step S22: The roadside unit calculates the global optimization model M ITS And the main road models Constructing the first traffic load optimization model in, Optimizing models for the vehicle itself; To map RSU j The number of intelligent vehicles; u is the number of intelligent vehicles;

[0124] Step S23: The roadside unit sends the first traffic load optimization model To smart vehicles.

[0125] In this embodiment, the roadside unit is responsible for communicating with the intelligent vehicle with wireless communication, and the roadside unit receives the global optimization model M transmitted from the cloud network. ITS And the main road models And the global optimization model M ITS And the main road models Transferring the model solution to the smart vehicle can make full use of the existing computing resources of the smart vehicle, relieve the computing pressure of the cloud network, and improve the efficiency of computing resource utilization; represents the vehicle optimization model of the u-th vehicle, To map RSU j The number of intelligent vehicles, u is the number of intelligent vehicles, and the roadside unit constructs the first traffic load optimization model The first traffic load optimization model Represents the vehicle optimization model collected by the first intelligent vehicle to the uth intelligent vehicle on the jth roadside unit; the roadside unit sends the first traffic load optimization model To smart vehicles, provide the initial first traffic load optimization model for smart vehicles Intelligent vehicles can optimize the model according to the first traffic load The required parameters in are used to calculate the ego-vehicle model.

[0126] Furthermore, the intelligent vehicle receives the target trunk road optimization task and optimizes the target trunk road according to the global optimization model M ITS , main road models and the first traffic load optimization model Constructing the vehicle's own optimization model Upload the vehicle's own optimized model To the roadside unit, including:

[0127] Step S31: The intelligent vehicle receives the target trunk road optimization task issued by the cloud network;

[0128] Step S32: The intelligent vehicle receives the global optimization model M transmitted by the roadside unit ITS And the main road models

[0129] Step S33: The intelligent vehicle receives the first traffic load optimization model constructed by the roadside unit

[0130] Step S34: the intelligent vehicle initializes the local driving state of the intelligent vehicle and collects external environment information;

[0131] Step S35: the intelligent vehicle replans the destination navigation route;

[0132] Step S36: Establishing an intersection selection priority list and corresponding cost model (c turn ,c path ,c traffic );

[0133] Step S37: Construct the priority list and the corresponding cost model into the vehicle's own optimization model Upload to the roadside unit;

[0134] Among them, the intersection options include: left turn, straight ahead, right turn, cost model (c turn ,c path ,c traffic ) includes turning cost, distance cost, and predicted traffic cost of the selected road. (c turn ) is the cost model of the left turn cost, the distance cost, and the predicted traffic volume cost of the selected road, (c path ) is the cost model of the straight turning cost, the distance cost, and the predicted traffic volume cost of the selected road, (c traffic ) is a cost model that includes the turning cost of turning right, the distance cost, and the predicted traffic volume cost of the selected road.

[0135] In this embodiment, it is mainly composed of intelligent vehicles that can be connected to the Internet of Vehicles. The intelligent vehicles receive the target trunk road optimization task issued by the cloud network, start preparation work after receiving the optimization task, and receive the global optimization model M transmitted by the roadside unit. ITS , main road models The first traffic load optimization model constructed with roadside units In order to calculate the ego vehicle model according to the required parameters in the model; the intelligent vehicle initializes the ego vehicle state model, and continuously collects the information required for route planning near the road according to the various types of sensors and other equipment installed on it, including various types of single / binocular cameras, ultrasonic radars, lidars, and vehicle driving status (speed, acceleration, reference trajectory, etc.), and establishes the corresponding environment model; then establishes the priority list and loss cost model for the three road options of left turn, straight, and right turn (c turn ,c path ,c traffic), the cost model includes turning cost, distance cost, and predicted traffic cost of the selected road; the route is replanned based on the driving route navigation algorithm used by the vehicle itself, and the comprehensive optimal driving route under the three intersection road options of left turn, straight and right turn is calculated respectively. The optimal route aims to minimize the loss cost. According to the loss cost of the three road options, a priority list with the lowest cost and the corresponding cost model are established; the road selection priority list and the corresponding cost model are constructed as the vehicle's own optimization model Upload to the corresponding roadside unit via wireless connection. The choice of re-planned route is calculated locally in the smart vehicle, and the local privacy data of the smart vehicle is effectively protected by uploading the fork selection priority and the vehicle's own optimization model instead of the vehicle's destination and other information. During the transmission process, for the same smart vehicle, for each main road it passes through, the fork selection priority and local optimization model received by the roadside unit do not have a conversion relationship, but are only related to the current main road fork, further strengthening the ability to prevent malicious attackers from cracking user privacy data.

[0136] Furthermore, the roadside unit receives the vehicle's own optimization model According to the first traffic load optimization model Integrate into the second traffic load optimization model Send the second traffic load optimization model To the edge computing unit, including:

[0137] Step S41: The roadside unit receives the vehicle self-optimization model established by the intelligent vehicle

[0138] Step S42: The roadside unit collects the driving status of non-intelligent vehicles

[0139] Step S43: The roadside unit optimizes the model based on the vehicle itself and non-intelligent vehicle driving status Integrate into the second traffic load optimization model

[0140] Step S44: The roadside unit sends the second traffic load optimization model To the edge computing unit.

[0141] In this embodiment, the roadside unit receives the vehicle optimization model established by the intelligent vehicle. Roadside units collect driving status of non-intelligent vehicles Including lane, speed, and optimizing model based on the vehicle itself and non-intelligent vehicle driving status Integrate into the second traffic load optimization model The second traffic load optimization model To the edge computing unit for accumulation and aggregation.

[0142] Further, the edge computing unit receives the second traffic load optimization model Solve the optimization results and establish the target trunk road congestion prediction model Send optimization results and target trunk road congestion prediction model To cloud networks, roadside units and smart road signs, including:

[0143] Step S51: The edge computing unit receives the global optimization model M sent by the cloud network ITS And the main road models

[0144] Step S52: The edge computing unit receives the second traffic load optimization model sent by the roadside unit And aggregate and accumulate to generate the second main road model

[0145] Step S53: The edge computing unit optimizes the global model M ITS , Vehicle self-optimization model And the second main road model Iteratively solve the optimization results;

[0146] Step S54: Establishing a target trunk road congestion prediction model based on the optimization results

[0147] Step S55: The edge computing unit sends the optimization results and the target trunk road congestion prediction model to cloud networks, roadside units and smart road signs.

[0148] In this embodiment, the edge computing unit receives the initialized global optimization model M sent by the cloud network. ITS And initialize the main road model The edge computing unit receives the second traffic load optimization model sent by the roadside unit The second traffic load optimization model Represents the set of vehicle optimization models collected by the jth roadside unit in the area; and aggregates and accumulates to generate the second main road model The second main road model Represents the set of traffic load optimization models collected by all roadside units on the i-th trunk road in the area; iterates to solve the optimization result, which includes one of the results of left turn, right turn, and straight driving. The edge computing unit establishes the target trunk road congestion prediction model based on the optimization result Including the target trunk road throughput within the cycle; finally, the optimization results and the target trunk road congestion prediction model are combined It is sent to the cloud network, roadside units and smart road signs, only considering the traffic load optimization of each main road, and assigning the optimization solution task to the edge computing unit to reduce the computing pressure of the cloud network.

[0149] Preferably, the target trunk road congestion prediction model Passed to the roadside unit, considering that in the future the roadside unit can send decision suggestions for non-intelligent vehicles.

[0150] Furthermore, the edge computing unit optimizes the global model M ITS , Vehicle self-optimization model And the second main road model Iterative solution optimization results, including:

[0151] Step S531: Initialize the global optimization model M ITS and load optimization models;

[0152] Step S532: constructing an intelligent vehicle fork junction selection model;

[0153] Step S533: Select the intelligent vehicle fork with the highest priority as the initial solution;

[0154] Step S534: adjusting the selection priority list of the intelligent vehicle according to the fork capacity;

[0155] Step S535: Iterate step S534 until the algorithm converges to the minimum cost that meets the traffic capacity of the fork intersection and obtains the optimization result.

[0156] In this embodiment, the present invention adopts a delayed receiving algorithm to dispatch intelligent vehicles with smaller cost losses to update road selection, initialize the traffic capacity model and load optimization model of the main road fork; take the current road selection of each intelligent vehicle as the initial solution X, update the intelligent vehicle road selection priority list and loss cost model, select the vehicle with the smallest loss cost caused by changing the road selection for road update, update the global load optimization result, and check the maximum passable vehicle capacity of the target main road; if the global load optimization result or the maximum passable vehicle capacity of the target main road does not meet the target threshold, repeat the update of the global load optimization result, and if the algorithm converges ahead of time, transfer the load optimization result; if the number of iterations reaches the maximum value, stop the above steps.

[0157] Furthermore, the edge computing unit sends the optimization results and the target trunk road congestion prediction model To cloud networks, roadside units and smart road signs, including:

[0158] Step S551: The edge computing unit sends the optimization results and the target trunk road congestion prediction model To cloud network and roadside units;

[0159] Step S552: The edge computing unit uses the target trunk road congestion prediction model Delivered to smart road signs;

[0160] Step S553: ​​Displaying the road congestion situation on the intelligent road sign.

[0161] In this embodiment, the edge computing unit sends the optimization results and the target trunk road congestion prediction model The optimization results are sent to the edge computing unit through the cloud network. The edge computing unit transmits the optimization results to the smart vehicle through the roadside unit, alleviating the delay problem of the vehicle receiving the optimization results from the cloud network. The smart vehicle chooses whether to make a scheduling decision based on the global optimization results aggregated by the cloud network and the judgment results of the target threshold. The edge computing unit transmits the target trunk road congestion prediction model to the smart vehicle. The information is transmitted to the smart road sign, and the road congestion situation is displayed on the smart road sign. Thus, the road congestion situation provides a suggestion for the road selection of the smart vehicle. The smart vehicle can choose the road according to the road congestion situation displayed on the smart road sign, so as to further alleviate the congestion of the main road.

[0162] Furthermore, the cloud network receives the optimization results and aggregates them into a global optimization result, compares the global optimization result with the target threshold, and if the global optimization result does not reach the target threshold, repeats steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision, including:

[0163] Step S61: The cloud network receives the optimization results and aggregates them into a global optimization result;

[0164] Step S62: Compare the global optimization result with the target threshold;

[0165] Step S63: Based on the comparison between the global optimization result and the target threshold, an intelligent vehicle dispatching decision is made or the instructions of steps S2-S5 are repeated. If the global optimization result does not reach the target threshold, the cloud network gives an instruction to repeat steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle dispatching decision.

[0166] In this embodiment, the cloud network receives the optimization results of the edge computing unit optimization solution and aggregates them into a global optimization result in the cloud network. According to the comparison between the global optimization result and the target threshold, the cloud network makes an instruction. If the global optimization result meets the target threshold, the intelligent vehicle receives the optimization result transmitted by the roadside unit and makes an intelligent vehicle scheduling decision. If the global optimization result does not meet the target threshold, the cloud network issues a repeat instruction and repeats steps S2-S5.

[0167] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A vehicle dispatching method for optimizing traffic load on main roads at intersections, characterized in that: The vehicle dispatching method comprises: Step S1: The cloud network publishes the target trunk road optimization task and builds a global optimization model And the main road models , where i is the number of a target trunk road, and P is the number of trunk roads; Step S2: The roadside unit constructs the first traffic load optimization model , sending the first traffic load optimization model To the intelligent vehicle, where j is the number of a roadside unit; Step S3: The intelligent vehicle receives the target trunk road optimization task and calculates the target trunk road optimization task according to the global optimization model. , the main road models and the first traffic load optimization model , build the vehicle's own optimization model , upload the vehicle's own optimization model To the roadside unit, where u is the number of smart vehicles; Step S4: The roadside unit receives the vehicle's own optimization model , according to the first traffic load optimization model Integrate into the second traffic load optimization model , send the second traffic load optimization model to the edge computing unit; Step S5: The edge computing unit receives the second traffic load optimization model , solve the optimization results and establish the target trunk road congestion prediction model , sending the optimization result and the target trunk road congestion prediction model To the cloud network, the roadside unit and the smart road sign; Step S6: the cloud network receives the optimization results and aggregates them into a global optimization result, compares the global optimization result with the target threshold, and if the global optimization result does not reach the target threshold, repeats steps S2-S5; if the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision; Among them, the second traffic load optimization model The first traffic load optimization model containing specific values ​​after integration .

2. The vehicle dispatching method according to claim 1, characterized in that: The cloud network publishes the target trunk road optimization task and builds a global optimization model And the main road models ,include: Step S11: The cloud network constructs the global optimization model And the main road models , and initialize the global optimization model And the main road models ; Constructing the global optimization model ; Construct the main road models ; in, is the number of roadside unit mappings corresponding to the main road i; P is the number of main road sets; i is the number of a target main road, the total number is I; j is the number of a roadside unit, the total number is J; Step S12: the cloud network publishes the target trunk road optimization task to the roadside unit, the edge computing unit and the smart vehicle; Step S13: Send the global optimization model And the main road models To the roadside unit and the edge computing unit, and allocate corresponding main roads to the roadside unit and the edge computing unit.

3. The vehicle dispatching method according to claim 2, wherein the global optimization model is sent And the main road models To the roadside unit and the edge computing unit, and assigning corresponding trunk roads to the roadside unit and the edge computing unit, including: Step S130: The cloud network sends the global optimization model And the main road models To the roadside unit and the edge computing unit; Step S131: Initialize the number of cluster centers and the locations of cluster centers; Step S132: Calculate the similarity between each main road and the cluster center, classify the main roads into the most similar class, and allocate cluster groups according to the principle of the shortest transmission distance; Step S133: updating the cluster center; Step S134: if the cluster centers have not converged, repeat steps S132-S133; Step S135: establishing an optimization demand matrix for each cluster; Step S136: Initializing the supply matrix of the edge computing unit for the optimization task; Step S137: replanning the demand matrix of each cluster, and adjusting the clustering of some trunk road optimization tasks; Step S138: If the supply of the edge computing unit does not satisfy the demand matrix, repeat step S137; Step S139: Establish an edge computing selection model and allocate the edge computing units to each cluster group.

4. The vehicle dispatching method according to claim 1, characterized in that: Roadside unit builds the first traffic load optimization model , sending the first traffic load optimization model To smart vehicles, including: Step S21: The roadside unit receives the global optimization model And the main road models , the global optimization model And the main road models transmitting to the smart vehicle; Step S22: The roadside unit calculates the global optimization model And the main road models , construct the first traffic load optimization model ;in, Optimizing models for the vehicle itself; For mapping The number of intelligent vehicles; u is the number of intelligent vehicles; Step S23: The roadside unit sends the first traffic load optimization model To the smart vehicle.

5. The vehicle dispatching method according to claim 1, characterized in that: The intelligent vehicle receives the target trunk road optimization task, and according to the global optimization model , the main road models and the first traffic load optimization model , build the vehicle's own optimization model , upload the vehicle's own optimization model To the roadside unit, including: Step S31: the intelligent vehicle receives the target trunk road optimization task issued by the cloud network; Step S32: The intelligent vehicle receives the global optimization model transmitted by the roadside unit And the main road models ; Step S33: The intelligent vehicle receives the first traffic load optimization model constructed by the roadside unit ; Step S34: the smart vehicle initializes the local driving state of the smart vehicle and collects external environment information; Step S35: the intelligent vehicle replans the destination navigation route; Step S36: Establishing an intersection selection priority list and corresponding cost model ; Step S37: Construct the priority list and the corresponding cost model into a vehicle optimization model , uploaded to the roadside unit; The intersection options include: left turn, straight ahead, right turn, and the cost model Including turning cost, distance cost, and predicted traffic cost of the selected road. ) is the cost model of the left turn cost, the distance cost, and the predicted traffic volume cost of the selected road, ) is the cost model for the straight-line turning cost, distance cost, and predicted traffic volume cost of the selected road. ) is the cost model for the right turn cost, distance cost, and predicted traffic volume cost of the selected road.

6. The vehicle dispatching method according to claim 1, characterized in that: The roadside unit receives the vehicle's own optimization model , according to the first traffic load optimization model Integrate into the second traffic load optimization model , send the second traffic load optimization model To the edge computing unit, including: Step S41: The roadside unit receives the vehicle self-optimization model established by the intelligent vehicle ; Step S42: The roadside unit collects the driving status of non-intelligent vehicles ; Step S43: The roadside unit optimizes the model based on the vehicle itself and the driving state of the non-intelligent vehicle , integrated into the second traffic load optimization model ; Step S44: The roadside unit sends the second traffic load optimization model To the edge computing unit.

7. The vehicle dispatching method according to claim 1, characterized in that: The edge computing unit receives the second traffic load optimization model , solve the optimization results and establish the target trunk road congestion prediction model , sending the optimization result and the target trunk road congestion prediction model To the cloud network, the roadside unit and the smart road sign, including: Step S51: The edge computing unit receives the global optimization model sent by the cloud network And the main road models ; Step S52: The edge computing unit receives the second traffic load optimization model sent by the roadside unit , and aggregate and accumulate to generate the second main road model ; Step S53: The edge computing unit calculates the value of the global optimization model according to the global optimization model. , the vehicle's own optimization model And the second main road model Iteratively solve the optimization results; Step S54: Establishing a target trunk road congestion prediction model based on the optimization results ; Step S55: The edge computing unit sends the optimization result and the target trunk road congestion prediction model To the cloud network, the roadside unit and the smart road sign.

8. The vehicle dispatching method according to claim 7, characterized in that: The edge computing unit is based on the global optimization model , the vehicle's own optimization model And the second main road model Iterative solution optimization results, including: Step S531: Initialize the global optimization model and load optimization models; Step S532: constructing an intelligent vehicle fork junction selection model; Step S533: selecting the intelligent vehicle fork with the highest priority as the initial solution; Step S534: adjusting the selection priority list of the intelligent vehicle according to the fork capacity; Step S535: iterate the step S534 until the algorithm converges to the minimum cost that meets the traffic capacity of the fork intersection, and obtains the optimization result.

9. The vehicle dispatching method according to claim 7, characterized in that: The edge computing unit sends the optimization result and the target trunk road congestion prediction model To the cloud network, the roadside unit and the smart road sign, including: Step S551: The edge computing unit sends the optimization result and the target trunk road congestion prediction model to the cloud network and the roadside unit; Step S552: The edge computing unit uses the target trunk road congestion prediction model Transmitting to the smart road sign; Step S553: ​​The intelligent road sign displays the road congestion situation.

10. The vehicle dispatching method according to claim 1, characterized in that: The cloud network receives the optimization results and aggregates them into a global optimization result, compares the global optimization result with the target threshold, and if the global optimization result does not reach the target threshold, repeats steps S2-S5; if the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision, including: Step S61: the cloud network receives the optimization results and aggregates them into a global optimization result; Step S62: comparing the global optimization result with the target threshold; Step S63: According to the comparison result between the global optimization result and the target threshold, an intelligent vehicle scheduling decision is given or the instructions of steps S2-S5 are repeated. If the global optimization result does not reach the target threshold, the cloud network gives an instruction to repeat steps S2-S5. If the global optimization result reaches the target threshold, the cloud network gives an intelligent vehicle scheduling decision.

Images