Unmanned aerial vehicle-based expressway traffic flow induction method and system

Abstract

The invention discloses an expressway traffic flow induction method and system based on an unmanned aerial vehicle. The system comprises a data acquisition and processing subsystem, a cloud traffic flow simulation subsystem and an induction subsystem. According to the method, control areas are divided for frequently-congested road sections (ramps and service areas) and upstream areas of the expressway, data acquisition is carried out, cloud simulation models of the corresponding control areas are constructed, and meanwhile, real-time updating of the models is carried out according to a certain time window; after the system recognizes congestion, a speed limiting strategy is calculated through the cloud model, the speed limiting strategy is sent to the airport, the unmanned aerial vehicle departs from the designated airport, arrives at the designated position of the road section and carries out speed limiting prompting to carry out traffic flow induction, the effect of relieving the traffic condition of the congested road section is achieved, and meanwhile the system can be used for accidental accident prompting.

Description

Technical Field

[0001] The present invention relates to a highway traffic flow induction method, and in particular to a highway traffic flow induction method and system based on unmanned aerial vehicle. Background Art

[0002] With the increase in the number of cars, traffic congestion on highways is becoming increasingly serious, especially near special sections (ramp entrances, service areas). In a certain period of time, due to the concentration of traffic flow, serious congestion often occurs. Traditional traffic flow induction systems often react slowly and cannot respond to sudden traffic conditions in real time. However, drones, due to their maneuverability and flexibility, can perform speed limit control at any location, which can become a new option for solving such problems. Patent CN202311188583 discloses a traffic control system and control method for preventing traffic congestion, including a road anomaly unit, a climate anomaly unit and a historical comparison unit. The road anomaly unit reports information on urban road maintenance and road repairs online. When maintenance and construction are carried out near the intersection of the expressway underpass, the maintenance and construction information will be sent and reported. By collecting information on climate and road conditions in advance, when it is judged that the intersection of the underpass will be congested in the later period after collection, the motor vehicles parked near the underpass intersection are evacuated through the road condition preprocessing unit. However, this solution can only prevent static factors that cause traffic congestion, such as road maintenance, and cannot respond promptly to sudden and occasional dynamic factors, such as traffic accidents. In addition, this solution can only provide warnings to vehicles crossing the road, and cannot guarantee the induction effect, so it has defects. Summary of the invention

[0003] Purpose of the invention: In view of the above problems, the present invention proposes a highway traffic flow induction method and system based on drones, which can ensure the real-time performance of the road induction system, improve the timeliness of the induction, and at the same time improve the induction effect.

[0004] Technical solution: The technical solution adopted by the present invention is a highway traffic flow induction method based on drones, comprising the following steps:

[0005] S1. The ramp or service area to the upstream main highway toll station is designated as the control area, and the control area is divided into three control sections: congestion monitoring area, traffic flow induction area and toll station area;

[0006] S2. Construct a traffic flow simulation model of the control area based on the road geometry data and historical traffic flow data of the control area; obtain the current real-time traffic flow information of the control area within the time window t, upload it to the traffic flow simulation model, and update the traffic flow simulation model in real time; the traffic flow simulation model outputs the traffic status information of each location in the control area;

[0007] S3. Determine the real-time congestion status of the congestion monitoring area according to the traffic status information of each location in the control area; predict the traffic status information of the congestion monitoring area through the traffic flow simulation model according to the traffic status information of each location in the control area and the real-time traffic flow data of the toll station area, and then predict the congestion status of the congestion monitoring area in the future according to the predicted traffic status information of the congestion monitoring area;

[0008] S4. When a congestion condition is detected or predicted in a congestion monitoring area, a speed limit strategy for a traffic flow induction area is calculated by a machine learning algorithm;

[0009] S5. Send the speed limit strategy task to the drone airport control center. The drone belonging to the airport will take off and cruise to the mission target location before issuing a speed limit warning.

[0010] The judgment of the congestion condition in step S3 includes: controlling the traffic status information of each position in the area, calculating the average vehicle speed and average headway in the traffic flow induction area; and judging the congestion condition by using a threshold method based on the average vehicle speed and average headway.

[0011] The speed limit strategy of the traffic flow induction area is calculated by the machine learning algorithm described in step S4, and a reinforcement learning algorithm is adopted. The environmental state quantity of the reinforcement learning algorithm includes the traffic state information of each position in the control area, and the action space is the speed limit value of each position in the traffic flow induction area. The reward function is negatively correlated with the traffic congestion condition in the congestion monitoring area and the traffic pressure caused by the traffic flow of the upstream toll station on the traffic flow induction area. The more serious the congestion condition, the larger the traffic flow, the greater the traffic pressure, and the lower the reward value.

[0012] The reinforcement learning algorithm is used to calculate the speed limit strategy in the traffic flow induction area, which includes the following steps:

[0013] S41, obtaining traffic status information of each location in the control area through a real-time updated traffic flow simulation model;

[0014] S42, initializing a reinforcement learning environment, an action space, and a reward function, wherein the environment state includes traffic state information of each location in the control area obtained by a real-time updated traffic flow simulation model;

[0015] S43, using the deep Q network DQN to train the agent, calculate the action and the corresponding reward function value according to the real-time traffic status, and select the optimal speed limit position and speed limit value according to the reward function value;

[0016] S44, inputting the speed limit position and speed limit value decided by the intelligent agent into the traffic flow simulation model, obtaining feedback from the traffic flow simulation model, and updating the intelligent agent;

[0017] S45. Repeat steps S43-S44 until the model converges and output the speed limit strategy corresponding to the current agent.

[0018] The specific form of the reward function is:

[0019] R t = -α·T congestion +β·V average -γ·P pressure

[0020] Among them, T congestion is the congestion time of the congestion monitoring area, V average is the average speed of traffic in the traffic flow induction area, P pressure represents the traffic pressure caused by the traffic volume of the upstream toll station on the traffic flow induction area, α, β, γ are the corresponding weight coefficients;

[0021]

[0022] D delay represents the upstream delay, which is the additional waiting time caused by congestion or speed limit when vehicles are queuing to enter the toll booth; V input The traffic volume in the upstream toll station area;

[0023] D delay =T actual -T expected

[0024] Where T actual is the time required for a vehicle to actually pass a certain distance R near the toll station, T expected is the estimated travel time to travel route R without congestion.

[0025] The agent is trained using the deep Q network DQN. The Q value update formula during training is:

[0026]

[0027] Among them, Q represents the current state S t Next, perform action a t Q value, S t and S t+1 Represent the current state and the next state respectively, a t and a t+1 Represents the currently executed action and the next action, R t Indicates that at time step t, the agent is in state S t Execute action a t The immediate reward obtained after, α0 is the learning rate, γ0 is the discount factor, Q target The Q value calculated for the target network.

[0028] The airport's drone takes off after receiving the mission and cruises to the mission target location to issue a speed limit warning, including the following steps:

[0029] S51. The airport receives the speed limit policy and performs task assignment, airspace allocation, and path planning for idle drones.

[0030] S52. The UAV accepts the mission, goes through the takeoff decision process, and rises to the target altitude; the UAV cruises to the target location according to the preset path planning.

[0031] S53. The UAV arrives at the mission location, automatically calibrates its position, lands above the highway roadside guardrail, and is captured and fixed; the UAV's guidance information screen rises and displays the speed limit information.

[0032] The present invention proposes a highway traffic flow induction system based on unmanned aerial vehicles, comprising a data acquisition and processing subsystem, a cloud-based traffic flow simulation subsystem memory and an induction subsystem, wherein the cloud-based traffic flow simulation subsystem comprises a processor and a computer program stored in the memory and executable on the processor, wherein the data acquisition and processing subsystem comprises three data acquisition modules, wherein the first data acquisition module is arranged near the ramp entrance or service area in a congestion monitoring area, the second data acquisition module is arranged in a section of a traffic flow induction area, and the third data acquisition module is arranged near a main toll station; when the processor in the cloud-based traffic flow simulation subsystem executes the computer program, the highway traffic flow induction method based on unmanned aerial vehicles is implemented.

[0033] The present invention proposes a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for inducing highway traffic flow based on a drone is implemented.

[0034] The present invention proposes a computer program product, including a computer program and / or instructions, which implement the highway traffic flow induction method based on unmanned aerial vehicle when executed by a processor.

[0035] Beneficial effects: Compared with the prior art, the present invention has the following advantages: the present invention divides the control area into the frequently congested sections (ramp entrances, service areas) and upstream areas of the expressway, collects data, constructs a cloud simulation model of the corresponding control area, and updates the model in real time according to a certain time window, thereby ensuring the real-time performance of the induction system; at the same time, after the system recognizes the congestion, the speed limit strategy is calculated through the cloud model, and the speed limit strategy is sent to the airport. The drone departs from the designated airport and arrives at the designated location of the section, and performs speed limit prompts to induce traffic flow, thereby achieving the effect of alleviating the traffic conditions of the congested section. It can be used for occasional accident prompts, improves the timeliness of induction, and enhances the induction effect. The induction method and system of the present invention can not only deal with traffic congestion caused by static factors such as road maintenance, but also respond to sudden and occasional dynamic factors such as traffic accidents in a timely manner. The present invention realizes real-time monitoring and induction of highway traffic flow by utilizing the mobility and flexibility of drones. Through accurate speed limit prompts, congestion is effectively alleviated and the traffic efficiency of highways is improved. The system has broad application prospects and significant social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Specific division of road control areas;

[0037] Figure 2 A flow chart showing the calculation of the speed limit strategy in the traffic flow induction area of ​​the present invention;

[0038] Figure 3 This is a schematic diagram of the structure of the highway traffic flow induction system based on drones according to the present invention;

[0039] Figure 4 This is a schematic diagram of the overall workflow of the UAV-based highway traffic flow induction system described in the present invention. DETAILED DESCRIPTION

[0040] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0041] Example 1

[0042] The method for inducing highway traffic flow based on drones described in the present invention uses the mobility of drones to induce traffic flow (speed limit prompt) on highways based on data collection and traffic modeling. The method includes the following steps:

[0043] S1. The area from the ramp entrance (service area) to the upstream highway toll station is designated as the control area, and the control section is divided into the congestion monitoring area, the traffic flow induction area and the toll station area.

[0044] According to the different locations of highway toll stations, they can be divided into mainline toll stations and ramp toll stations. Mainline toll stations are generally located on the main line of the highway, with large traffic volume, fast speed and high traffic capacity requirements; ramp toll stations are located on the ramps or roads along the highway, with relatively small traffic volume. The target control area is divided on the main line of the highway.

[0045] Congestion monitoring area: The area 200 meters upstream of the ramp (service area) is divided into a congestion monitoring area, where cameras and geomagnetic induction coils are deployed to implement congestion identification and detection. The specific length of the congestion monitoring area is based on the historical traffic data of the road.

[0046] Toll station area: The toll station area is divided into the toll station area at the upstream of the ramp (service area) and 150 meters downstream. The area where vehicles accelerate after leaving the toll station is not subject to traffic flow induction measures. Use toll station information to obtain vehicle flow input data. The acceleration of the car during comfortable acceleration is 1m / s. According to regulations, when there are more than 3 lanes in the same direction, the minimum speed of the leftmost lane is 110 kilometers per hour, the minimum speed of the middle lane is 90 kilometers per hour, and the minimum speed of the rightmost lane is 60 kilometers per hour. The minimum acceleration distance of a vehicle leaving the toll station is about 138m, and 150m is taken as the distance of the acceleration area.

[0047] Traffic flow induction area: The road section between the toll station area and the congestion monitoring area is the main implementation area for drone speed limit tasks. By inducing and controlling the traffic flow in this area, the congestion in the downstream congestion monitoring area can be alleviated.

[0048] like Figure 1 The figure shows the specific division of the road control area, including the congestion monitoring area, the traffic flow induction area, and the toll station area. The first data acquisition module 1 of the congestion monitoring area is located near the ramp entrance (service area), and is used to collect the flow, average speed, average queue length, and headway near the ramp entrance (service area), including a video data acquisition unit and a geomagnetic coil detection auxiliary unit; the second data acquisition module 2 is located in the section of the traffic flow induction area, which is mainly a video data acquisition unit, and is used to collect the average speed of the traffic in the section of the traffic flow induction area; the third data acquisition module 3 is located at the entrance and exit of the toll station, and is a toll station gantry information unit, which is used to collect the unit time flow input status at the toll station.

[0049] S2. Obtain the road geometry data and historical traffic flow data of the control section to preliminarily build a traffic flow simulation model of the control section. Define a time window length, obtain the real-time traffic flow information of the current control section according to the time window t, upload it to the cloud simulation model, and update the simulation model in real time.

[0050] The traffic flow simulation model is modeled using SUMO. The traffic flow simulation model verifies the constructed simulation model according to certain evaluation methods and evaluation parameters. After the model verification is successful, the simulation model is reliable.

[0051] The specific steps for real-time updating of the simulation model are as follows:

[0052] S21. Obtain the road geometry data, historical traffic data, and average speed data of the control section, preliminarily build a traffic flow simulation model, and use simulation parameters such as traffic flow, average speed, vehicle parameters, and following model to verify the model to ensure the feasibility and reliability of the model.

[0053] S22. Using pre-installed cameras as the main data collection unit, through image recognition technology, obtain the vehicle type, flow, average speed, headway and other data in the control area (focusing on the congestion monitoring area). At the same time, use geomagnetic induction coils to collect relevant data as a supplement to the main data collection unit to repair missing data and correct abnormal data.

[0054] S23. Use the toll station gantry system to obtain the vehicle flow data input upstream of the current control area as the vehicle input of the simulation model.

[0055] S24, updating the data in the above steps according to the time window t, and inputting the simulation model to perform model updating.

[0056] S3. Based on the real-time traffic flow information in step S2, calculate the average traffic speed and average headway time in the congestion identification area, and judge the congestion situation based on the average speed and average headway time; at the same time, predict the congestion situation in the congestion monitoring area based on the flow input in the toll station area. The specific steps are as follows:

[0057] S31. Calculate the average speed and average headway of traffic flow using the collected traffic flow data. The specific formula is as follows:

[0058]

[0059] Get the speed v of the i-th vehicle within a certain period of time i , the number of vehicles is n.

[0060] h i =t i -t i-1

[0061]

[0062] The time series of vehicles passing a certain point t1, t2, t3, ..., t n , where t iis the time when the i-th vehicle passes a certain point, then the headway between the i-th vehicle and the i-1-th vehicle is h i =t i -t i-1 .

[0063] S32: The congestion identification method in the congestion monitoring area is the threshold method, which is used to identify congestion by setting the thresholds of the average speed of the traffic flow and the average headway. and When , it is determined to be in a congested state. Where v1 and v2 are the average speed of the traffic flow induction area and the average speed of the congestion monitoring area, respectively, which are calculated by the traffic flow information collected by the data acquisition unit 2 and the data acquisition unit 1. At the same time, the traffic conditions in the congestion identification area are predicted according to the upstream flow conditions.

[0064] S4. When congestion is detected or predicted in the congestion monitoring area, the speed limit strategy of the traffic flow induction area is calculated by machine learning using the real-time updated traffic flow simulation model. Figure 2 The figure shows the speed limit strategy calculation flow chart for the traffic flow induction area, which shows the specific process of data collection, state vector construction, model training and optimization, and real-time strategy generation. The specific steps for speed limit strategy generation are as follows:

[0065] S41. Obtain traffic status information of each location in the control area through a real-time updated traffic flow simulation model.

[0066] S42: Initialize a reinforcement learning environment, wherein the environment state includes traffic state information of each location in the control area.

[0067] S43. Define an action space and a reward function, where the action space is the speed limit values ​​at different positions in the traffic flow induction area, and the reward function is related to the traffic congestion condition in the ramp congestion monitoring area and the traffic pressure of the upstream toll station on the traffic flow induction area.

[0068] S44. Use the reinforcement learning algorithm deep Q network DQN to train the intelligent agent and select the optimal speed limit location and speed limit value according to the real-time traffic status.

[0069] S45. Apply the speed limit position and speed limit value decided by the intelligent agent to the simulation model, obtain simulation feedback, and update the intelligent agent.

[0070] S46. Repeat steps S4-S5 until the model converges, that is, the intelligent agent can effectively alleviate the congestion at the ramp entrance under different traffic conditions.

[0071] S47. Apply the trained intelligent agent to the actual traffic management system and adjust the speed limit strategy in real time.

[0072] The specific method for generating a speed limit policy is as follows:

[0073] 1. Environment modeling and state representation.

[0074] In the present invention, the environment is defined as a dynamic traffic system consisting of highway sections, vehicles, road infrastructure, weather conditions, and real-time traffic flows. t It is a comprehensive representation of the current traffic situation, including the following information:

[0075] Congestion status: reflects the traffic density, queue length or average vehicle speed in the congested area.

[0076] Average vehicle speed in the traffic flow induction area: the average driving speed in the current area affected by weather conditions.

[0077] Real-time traffic input status of upstream toll stations: vehicle flow data passing through high-speed toll stations.

[0078] Weather conditions: factors such as rain, snow, etc. that may affect driving safety and speed.

[0079] State vector S t It is composed of the above indicators and input into the deep Q network model for processing.

[0080] 2. Action definition.

[0081] The speed limit policy is implemented by action a t To express, the action set A contains all possible speed limit decisions. Each action a t It includes two dimensions:

[0082] Speed ​​limit location: A speed limit is imposed on a specific section of a traffic flow induction area at a certain distance from the congestion identification area.

[0083] Speed ​​limit: The specific speed limit set, such as 30km / h, 50km / h, 70km / h, 90km / h, etc.

[0084] In the present invention, the action space is designed as a discrete value set, covering multiple common speed limit strategy combinations.

[0085] 3. Reward function design.

[0086] Reward function R t Used to quantify the effectiveness of each speed limit strategy, the goal is to minimize traffic congestion by dynamically adjusting speed limit measures. The design of the reward function takes into account the following factors:

[0087] Congestion relief effect: measured by shortening the travel time of vehicles in the congestion identification area.

[0088] Traffic efficiency: quantified by increasing the average speed of traffic in the control area and reducing traffic congestion.

[0089] Safety: Consider the impact of weather conditions and other factors on safe driving, and avoid the risk of traffic accidents caused by excessive speed limits.

[0090] The specific form of the reward function can be expressed as:

[0091] R t = -α·T congestion +β·V average -γ·P pressure

[0092] Among them, T congestion is the congestion time of the congestion monitoring area, which is used to indicate the traffic congestion condition in the congestion monitoring area. It can be replaced by other quantities that can reflect the congestion condition. average is the average speed of traffic in the traffic flow induction area, P pressure It represents the traffic pressure caused by the flow input of the toll station on the downstream control area, and α, β, and γ are the corresponding weight coefficients.

[0093]

[0094] Upstream delay D delay It usually refers to the extra waiting time that vehicles have to wait in line to enter a toll booth due to congestion or speed limit, expressed in seconds (s) or minutes (min).

[0095] D delay =T actual -T expected

[0096] Where T actual is the time required for the vehicle to actually pass a certain distance, T expected It is the estimated travel time in an uncongested environment.

[0097] V input is the flow input condition of the upstream toll station area. The size of the flow input is proportional to the traffic pressure, that is, a larger flow input means a greater traffic pressure on the downstream control area. Therefore, it should be used as a negative penalty term in the reward function.

[0098] 4. Construction of deep Q network model.

[0099] The present invention uses a deep Q network (DQN) to construct a speed limit strategy generation model. The Q-network is a multi-layer neural network used to estimate state-action pairs (S t ,a t ), that is, in a given state S t Next, perform action a tThe future accumulated rewards that can be obtained include:

[0100] Input layer: receives the state vector S t as input.

[0101] Hidden layers: Several fully connected layers, using ReLU as the activation function.

[0102] Output layer: Output each possible action a t Q value.

[0103] In addition, in order to improve the stability of training, the present invention uses a target network (Target Network), that is, a Q network with the same structure as the main network but slower parameter update, to calculate the target Q value.

[0104] 5. Model training.

[0105] Model training is completed through the following steps:

[0106] Experience replay mechanism: Store the data of each interaction with the environment in the experience replay buffer, and randomly sample small batches of samples for training, thereby breaking the correlation between samples and reducing the instability of model training.

[0107] Q value update: Update based on the Bellman equation, that is:

[0108]

[0109] Among them, α is the learning rate, γ is the discount factor, and Q target The Q value calculated for the target network.

[0110] Exploration and Utilization Strategy: Using the ∈-greedy strategy, during the model training process, the agent selects random actions (exploration) with probability ∈, and selects the action with the highest current Q value (utilization) with probability 1-∈, thereby balancing the exploration and utilization of the strategy in an unknown environment.

[0111] 6. Generation and application of speed limit strategies.

[0112] Through the training of the above DQN model, the present invention can t Generate the optimal speed limit strategy a t The generated speed limit strategy contains specific speed limit locations and speed limit values, and is implemented through a dynamic traffic control system. The implementation effect is monitored through a feedback mechanism and continuously used for model optimization to improve the accuracy and effectiveness of speed limit decisions.

[0113] S5: Send the speed limit strategy task to the nearest drone airport. After receiving the task, the drone belonging to the airport takes off and cruises to the mission target location to issue a speed limit warning. The specific steps include:

[0114] S51. The airport receives the speed limit policy and performs task assignment, airspace allocation, and path planning for idle drones.

[0115] S52. The UAV accepts the mission, goes through the takeoff decision process, and rises to the target altitude; the UAV cruises to the target location according to the preset path planning.

[0116] S53. The UAV arrives at the mission location, automatically calibrates its position, lands above the highway roadside guardrail, and is captured and fixed; the UAV's guidance information screen rises and displays the speed limit information.

[0117] S6. After the speed limit strategy is implemented, real-time feedback on the strategy implementation is provided to monitor changes in congestion conditions in the congestion monitoring area. After several time windows, if the strategy implementation effect is not good, the strategy is updated and the drone changes the road speed limit warning.

[0118] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned drone-based highway traffic flow induction method is implemented.

[0119] In one embodiment, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the described UAV-based highway traffic flow induction method.

[0120] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0121] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0123] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0124] Example 2

[0125] The drone-based highway traffic flow induction system described in the present invention sets up a data acquisition system at frequently congested sections (usually ramp entrances), and uses, for example, cameras, geomagnetic induction coils, etc. to collect road flow information, and at the same time collects information on highway toll gates upstream of the ramp entrances, so as to provide effective data for traffic flow predictions in congested sections. At the same time, the traffic modeling system is used to model key sections, and the modeling is updated in real time based on the collected information. When congestion is identified, the speed limit location and speed limit information are calculated, and accurate speed limit prompts are given upstream of the congested section by drone, thereby alleviating congestion in the shortest time and improving traffic efficiency. Figure 3 The figure shows a schematic diagram of the structure of the highway traffic flow induction system based on drones according to the present invention, which includes a data collection and processing subsystem, a cloud traffic flow simulation subsystem and an induction subsystem.

[0126] Data acquisition and processing subsystem: including data acquisition module and congestion identification module; the data acquisition module includes video data acquisition unit, geomagnetic coil detection auxiliary unit, toll station gantry information unit. The video data acquisition unit is used to collect traffic flow information in the control area, including vehicle type, flow data, headway, etc. The geomagnetic coil detection auxiliary unit is used to repair missing data and correct abnormal data in the video data acquisition unit. The toll station gantry information unit is used to collect the traffic flow input from upstream to the control area. The congestion identification module uses an algorithm to calculate the average speed of the traffic flow and the average headway based on the traffic flow data collected by the data acquisition module, and uses the threshold method to identify and determine congestion.

[0127] Cloud traffic flow simulation subsystem: including simulation module and speed limit strategy module; the simulation module performs initial construction, model verification and model update of the traffic flow simulation model of the control area. The simulation system obtains the geometric data and historical traffic data of the control area, preliminarily constructs the traffic flow simulation model of the control area, and performs model correction to ensure the reliability of the simulation model; after the model correction is successful, the real-time traffic flow data of the control area uploaded by the data acquisition module is received, and the model is updated according to a certain time window. After receiving the congestion information, the speed limit strategy module calculates the optimal speed limit and the optimal speed limit position based on the simulation model and the special algorithm of machine learning with the goal of the shortest congestion duration. The optimal speed limit position is located at any position in the traffic flow induction area, and the minimum speed limit does not exceed the real-time congestion speed of the congestion monitoring area. After the drone performs the mission and implements the speed limit strategy, the simulation subsystem evaluates the implementation effect of the strategy through the data updated after several time windows. If the effect is not ideal, the speed limit strategy is updated; if the effect is ideal, the strategy is implemented until the congestion is eliminated.

[0128] Induction subsystem: including drone airport and drone module; the airport involves the layout planning of drone airport, the construction of airport facilities, etc. The goal of airport site selection is to enable drones to complete the implementation of tasks in multiple control areas within the minimum cruising range, which involves economic, efficiency, safety and other factors. The drone airport facilities include drone control system, drone energy system, air traffic management system, operation monitoring center, intelligence center, maintenance center, etc. The drone control system is responsible for controlling the automatic take-off and landing, autonomous cruising, mission execution, and return to the factory for recycling of drones; the drone energy system is responsible for the energy supply of drones to ensure the endurance of airport drones; the air traffic management system is mainly responsible for airspace division and drone cruising path planning; the operation monitoring center is responsible for monitoring the overall operation status of the airport and the safety of the drone operation process; the intelligence center is responsible for data transmission between the cloud large model and the airport, information exchange between various airport departments, and information transmission between drones and airports; the maintenance center is responsible for the maintenance and support of drones to ensure the stability and reliability of drones. The speed limit strategy module sends the speed limit strategy to the airport's intelligence center, and the intelligence center sends the speed limit strategy to various airport departments to complete the operational preparation work; the air traffic management department selects the appropriate altitude layer according to the strategy information and plans the UAV's cruise path; the UAV receives the airport information, goes through the takeoff decision process, rises to the target altitude, and takes off and cruises to the speed limit strategy target location according to the predetermined planned path; after precise positioning, it lands above the roadside guardrail of the highway to implement grabbing and fixing, raises the induction information display screen, displays the speed limit warning information, and reminds the driver to slow down in time to complete the traffic flow induction task. The UAV device system includes energy system, positioning system, flight control system, data transmission system, operating tools, and induction information display system; the navigation system and flight control system are responsible for the automatic cruising of the UAV according to the predetermined planned path, and the operating tools and induction information display system are responsible for the specific implementation of the induction task. After the UAV arrives above the designated location, the operating tool accurately identifies the roadside guardrail and implements the grabbing work to fix the UAV on the roadside guardrail; the data transmission system is responsible for receiving strategy update information.

[0129] The present invention can be applied to multiple ramps and congested sections of highways. By flexibly adjusting speed limit measures, it can effectively alleviate traffic congestion and improve the overall traffic flow efficiency. For example, the length of the "congestion monitoring area" of a highway ramp is 200 meters; the section between the ramp and the upstream highway toll station is divided into a "traffic flow induction area", which is 15 kilometers long.

[0130] like Figure 4 The figure shows the overall workflow of the highway traffic flow induction system based on drones according to the present invention. It includes the following processes:

[0131] Input relevant information to complete the system work preparation: the process inputs road geometry data and historical road traffic information to build an initial simulation model, determines the evaluation criteria, and calibrates the model with road traffic, average speed, vehicle parameters, etc. as parameters; through animation inspection and precision calibration, the model verification is successful and the model is reliable. After the model is built, the system enters the working state, and the data acquisition and processing subsystem transmits data to the cloud traffic flow subsystem according to a certain time window to update the model; the second data acquisition module 2 collects the traffic information in the traffic flow induction area and calculates the average speed, and the first data acquisition module 1 collects the traffic information in the congestion monitoring area. Some data is missing, and the data collected by the geomagnetic coil detection auxiliary unit is used to repair the data. At the same time, the congestion identification module identifies congestion; the toll station gantry information unit of the third data acquisition module 3 sends the past traffic information of the toll station to the cloud model for congestion prediction; the congestion identification module in the data acquisition and processing subsystem calculates and (where v1 and v2 are the average speed in the traffic flow induction area and the average speed in the congestion monitoring area, respectively, which are calculated by the traffic flow information collected by data acquisition unit 2 and data acquisition unit 1, respectively). The data acquisition and processing subsystem determines and identifies congestion, uploads the congestion information to the cloud simulation model subsystem, and the speed limit module runs the algorithm.

[0132] The flow input of the upstream toll station is 1500 vehicles / hour, and the current speed of the traffic flow induction area is 85km / h. According to the system settings, the system determines that it is in a congested state. After the system detects congestion, the simulation model calls the DQN algorithm in real time to generate a speed limit strategy. The state input of DQN includes the following variables: S t =[{congestion status},{upstream traffic input},{average speed},{weather conditions}]). Congestion status: congestion. Upstream traffic input: 1500 vehicles / hour. Average speed: 85km / h. Weather conditions: sunny. The DQN algorithm generates the following speed limit strategy based on the current status, historical data and traffic forecast: speed limit location: [speed limit 5 km upstream from the ramp: 90km / h].

[0133] The cloud traffic flow simulation subsystem selects the best airport and sends the generated speed limit strategy to the mission execution airport; the airport receives the mission information (including latitude and longitude and speed limit), selects an idle drone, allocates airspace, and plans the drone's flight path; the drone accepts the mission, rises to the target airspace after taking off, cruises to the mission location according to the predetermined path, arrives above the mission location, lowers its altitude, accurately locates above the roadside guardrail, grabs the fixed value on the guardrail, and raises the induction information display screen to display the speed limit information; according to the speed limit strategy requirements, the drone hangs a warning sign with a speed limit of 90 kilometers per hour 5 kilometers upstream of the congested section to remind drivers to gradually slow down, thereby reducing the overall speed of traffic upstream of the congested section; feedback after several time windows: 5 minutes after the speed limit strategy was implemented, the average speed in the congestion monitoring area returned to 70km / h, the upstream traffic volume stabilized at 1,400 vehicles / hour, and the congestion at the ramp entrance was alleviated. Strategy adjustment: If the system detects that the congestion continues, the speed limit value will continue to be adjusted until the congestion is completely alleviated.

[0134] This embodiment induces the speed of upstream traffic in the congested area through a precisely controlled speed limit strategy, thereby reducing the average speed of the traffic, avoiding congestion caused by traffic flow concentrated at the ramp entrance, and effectively alleviating road congestion.

Claims

1. A highway traffic flow induction method based on drones, characterized in that: The following steps are involved: S1. The ramp or service area to the upstream main highway toll station is designated as the control area, and the control area is divided into three control sections: congestion monitoring area, traffic flow induction area and toll station area; S2. Construct a traffic flow simulation model of the control area based on the road geometry data and historical traffic flow data of the control area; obtain the current real-time traffic flow information of the control area within the time window t, upload it to the traffic flow simulation model, and update the traffic flow simulation model in real time; the traffic flow simulation model outputs the traffic status information of each location in the control area; S3. Determine the real-time congestion status of the congestion monitoring area according to the traffic status information of each location in the control area; predict the traffic status information of the congestion monitoring area through the traffic flow simulation model according to the traffic status information of each location in the control area and the real-time traffic flow data of the toll station area, and then predict the congestion status of the congestion monitoring area in the future according to the predicted traffic status information of the congestion monitoring area; S4. When a congestion condition is detected or predicted in a congestion monitoring area, a speed limit strategy for a traffic flow induction area is calculated by a machine learning algorithm; S5. Send the speed limit strategy task to the drone airport control center. The drone belonging to the airport will take off and cruise to the mission target location before issuing a speed limit warning.

2. The method for inducing highway traffic flow based on drone according to claim 1, characterized in that: The judgment of the congestion condition in step S3 includes: controlling the traffic status information of each position in the area, calculating the average vehicle speed and average headway in the traffic flow induction area; and judging the congestion condition by using a threshold method based on the average vehicle speed and average headway.

3. The method for inducing highway traffic flow based on drone according to claim 1, characterized in that: The speed limit strategy of the traffic flow induction area is calculated by the machine learning algorithm described in step S4, and a reinforcement learning algorithm is adopted. The environmental state quantity of the reinforcement learning algorithm includes the traffic state information of each position in the control area, and the action space is the speed limit value of each position in the traffic flow induction area. The reward function is negatively correlated with the traffic congestion condition in the congestion monitoring area and the traffic pressure caused by the traffic flow of the upstream toll station on the traffic flow induction area. The more serious the congestion condition, the larger the traffic flow, the greater the traffic pressure, and the lower the reward value.

4. The method for inducing highway traffic flow based on drone according to claim 3 is characterized in that: The reinforcement learning algorithm is used to calculate the speed limit strategy in the traffic flow induction area, which includes the following steps: S41, obtaining traffic status information of each location in the control area through a real-time updated traffic flow simulation model; S42, initializing a reinforcement learning environment, an action space, and a reward function, wherein the environment state includes traffic state information of each location in the control area obtained by a real-time updated traffic flow simulation model; S43, using the deep Q network DQN to train the agent, calculate the action and the corresponding reward function value according to the real-time traffic status, and select the optimal speed limit position and speed limit value according to the reward function value; S44, inputting the speed limit position and speed limit value decided by the intelligent agent into the traffic flow simulation model, obtaining feedback from the traffic flow simulation model, and updating the intelligent agent; S45. Repeat steps S43-S44 until the model converges and output the speed limit strategy corresponding to the current agent.

5. The method for inducing highway traffic flow based on drone according to claim 4, characterized in that: The specific form of the reward function is: R t =-α·T congestion +β·V average -γ·P pressure Among them, T congestion is the congestion time of the congestion monitoring area, V average is the average speed of traffic in the traffic flow induction area, P pressure represents the traffic pressure caused by the traffic volume of the upstream toll station on the traffic flow induction area, α, β, γ are the corresponding weight coefficients; D delay represents the upstream delay, which is the additional waiting time caused by congestion or speed limit on the road section where vehicles queue to enter the toll station; V input The traffic volume in the upstream toll station area; D delay =T actual -T expected Where T actual is the time required for the vehicle to actually pass a certain distance R near the toll station, T expected is the estimated travel time to travel route R without congestion.

6. The method for inducing highway traffic flow based on drone according to claim 4, characterized in that: The agent is trained using the deep Q network DQN. The Q value update formula during training is: Among them, Q represents the current state S t Next, perform action a t Q value, S t and S t+1 Represent the current state and the next state respectively, a t and a t+1 Represents the currently executed action and the next action, R t Indicates that at time step t, the agent is in state S t Execute action a t The immediate reward obtained after, α0 is the learning rate, γ0 is the discount factor, Q target The Q value calculated for the target network.

7. The method for inducing highway traffic flow based on drone according to claim 1, characterized in that: The airport's drone takes off after receiving the mission and cruises to the mission target location to issue a speed limit warning, including the following steps: S51. The airport receives the speed limit policy and performs task assignment, airspace allocation, and path planning for idle drones. S52. The UAV accepts the mission, goes through the takeoff decision process, and rises to the target altitude; the UAV cruises to the target location according to the preset path planning. S53. The UAV arrives at the mission location, automatically calibrates its position, lands above the highway roadside guardrail, and is captured and fixed; the UAV's guidance information screen rises and displays the speed limit information.

8. A highway traffic flow induction system based on drones, comprising a data acquisition and processing subsystem, a cloud traffic flow simulation subsystem memory and an induction subsystem, wherein the cloud traffic flow simulation subsystem comprises a processor and a computer program stored in the memory and executable on the processor, characterized in that: The data acquisition and processing subsystem includes three data acquisition modules, the first data acquisition module is arranged near the ramp entrance or service area in the congestion monitoring area, the second data acquisition module is arranged in the section of the traffic flow induction area, and the third data acquisition module is arranged near the main toll station; when the processor in the cloud-based traffic flow simulation subsystem executes the computer program, the drone-based highway traffic flow induction method described in any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the highway traffic flow induction method based on a drone according to any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or instructions are executed by a processor, the highway traffic flow induction method based on a drone as described in any one of claims 1 to 7 is implemented.

Images