A method for constructing a city road network simulation model for parking behavior experiments

Abstract

The application relates to a kind of urban road network simulation model construction methods for parking behavior experiment, comprising the following steps: S1, establishing traffic simulation system, simulating urban road network traffic flow and curb parking system;S2, establish simulation driving system, realize driving environment simulation, and provide real-time navigation of road condition, complete parking location simulation experiment in the first perspective;S3, establish data acquisition and analysis system, the trajectory data and global traffic information data of parking location simulation experiment are collected in real time screening, and are converted into expert trajectory data set input multiple algorithms, carry out the processing and model training of data, obtain strategy function;S4, vehicle information and global traffic information are solved by strategy function, form a single action instruction, guide the next step movement strategy of vehicle.Compared with prior art, the application has the advantages of close to reality, low data delay, providing support for subsequent research and the like.

Description

Technical Field

[0001] This invention relates to the field of traffic simulation, and in particular to a method for constructing a simulation model of an urban road network for parking behavior experiments. Background Technology

[0002] In response to the parking problem, a number of optimization measures for urban parking resource allocation and road network traffic operation management have been introduced. Currently, China faces problems such as insufficient parking supply capacity, low governance levels, and lagging marketization, necessitating the optimization of parking information management, the promotion of intelligent parking services, and encouragement of parking resource sharing.

[0003] Because existing traffic system analysis theories lack a synergistic consideration of parking behavior and traffic conditions, current measures are ineffective, directly hindering cities from further improving their quality and management services. At the individual traveler level, there is a lack of refined and personalized parking guidance schemes, making it difficult for travelers to effectively assess the impact of their parking behavior on themselves and the entire parking system. Furthermore, for the parking system as a whole, there is a lack of statistical indicators and modeling optimization schemes. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method for constructing an urban road network simulation model for parking behavior experiments. By constructing an urban road network simulation model, the method reflects the different parking location decision-making behaviors of users in different traffic environments at the micro level. At the same time, it explores the interaction and mutual influence mechanism between parking location behavior and road traffic conditions, and considers the role of parking behavior and traffic conditions in the parking system.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A method for constructing an urban road network simulation model for parking behavior experiments, the method comprising the following steps:

[0007] S1. Establish a traffic simulation system to simulate urban road network traffic flow and roadside parking system;

[0008] S2. Establish a driving simulation system to simulate the driving environment in the traffic simulation system in S1 and provide real-time road condition navigation. Complete the parking space finding simulation experiment from a first-person perspective.

[0009] S3. Establish a data acquisition and analysis system to collect and filter the trajectory data and global traffic information data of the parking space search simulation experiment in S2 in real time, and convert them into expert trajectory datasets. Input the expert trajectory datasets into various algorithms to process the data and train the models to obtain the policy functions.

[0010] S4. Solve the vehicle information and global traffic information through the strategy function to form a single action command to guide the vehicle's next movement strategy.

[0011] Furthermore, step S1 specifically includes:

[0012] S101. Select the target road network, use the game engine to model it, extract the road network information into a graph structure composed of points and edges, and store urban traffic rule information with points and edges as the basic units.

[0013] S102. The simulated urban road network traffic flow is abstracted into specific vehicles, and vehicle information is assigned to them. A complete urban road network traffic flow is established through the dynamic interaction between vehicles and road network nodes.

[0014] S103. Taking road segments as the basic operation unit, and road segment feature information as traffic rule information and real-time traffic flow information of urban road network, a logical subject is established as the main traffic rule carrier, giving road segments an executable operation space.

[0015] S104. Assign a pathfinding algorithm to vehicle units that simulate urban road network traffic flow to form diverse and complex urban road network traffic flow.

[0016] S105. To enable vehicle units that simulate urban road network traffic flow to detect obstacles around the vehicle, an arc-shaped detection surface is created based on the ray detection mechanism of the game engine. The ray density, detection angle range, and number of ray layers in the control are controlled by parameter settings.

[0017] Furthermore, the vehicle information in step S102 includes vehicle location information, vehicle speed information, and vehicle volume information.

[0018] Furthermore, the pathfinding algorithm in step S104 includes a global optimal path selection algorithm, a local optimal path selection algorithm, and a path selection algorithm based on specific population preferences; the path selection algorithm based on specific population preferences uses reinforcement learning to obtain a policy function and calculates the path selection parameters frame by frame.

[0019] Furthermore, step S2 specifically includes:

[0020] S201. Construct a basic acceleration control framework to enable the driver to control the vehicle's direction and speed via keyboard operation.

[0021] S202. After repeated real-person experiments and adjustments to the driver and vehicle driving-related variables, the vehicle driving is made closer to the real situation.

[0022] S203. Based on the existing physical entity simulation function of the game engine, realize the detection of collisions between the driver vehicle and objects in the traffic simulation system established in S1, as well as the handling of special boundary cases.

[0023] S204. The driver's vehicle is equipped with an automatic following function for the game engine navigation camera;

[0024] S205. By using layered visualization, the location of the vehicle driven by the experimenter is magnified on the map, and the traffic flow information on the map is updated in real time to achieve automatic guidance and calculation of the target location.

[0025] Furthermore, the keyboard operation in step S201 is as follows: the W key imparts forward acceleration to the vehicle, the S key imparts acceleration to the vehicle in the opposite direction to the current direction of the vehicle's direction, the A key controls the vehicle to turn left, and the D key controls the vehicle to turn right.

[0026] Furthermore, step S3 specifically includes:

[0027] S301. Obtain real-time global traffic information through urban road network traffic flow, perform data correctness detection based on traffic rules, and preprocess the data to form a data test set with trajectory as the basic data format.

[0028] S302. By matching the coordinates of all vehicles in the experiment with the coordinate range of road segments and parking spaces, the real-time traffic flow and parking information of each road segment are statistically obtained, thereby constructing global feature information.

[0029] S303. By matching the real-time coordinates of the vehicle controlled by the experimenter with the road segment coordinate range, the location information of the controlled vehicle on the road segment is obtained in real time, and the information is saved to construct an expert trajectory dataset.

[0030] S304. Input the obtained trajectory data into various algorithms for model training to obtain the policy function.

[0031] Furthermore, the various algorithms mentioned include clustering algorithms, inverse reinforcement learning algorithms, and reinforcement learning algorithms.

[0032] Furthermore, step S4 specifically involves storing the entire strategy function in a background file. When a vehicle in the traffic simulation stream arrives at the decision intersection, the current vehicle information and global traffic information are automatically input into the strategy function for solving. The resulting solution forms a single action command to guide the vehicle's next movement strategy.

[0033] Furthermore, the action commands include straight-ahead commands, left-turn commands, right-turn commands, and U-turn commands.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] I. This invention uses the high-performance game engine Unity to construct a city road network simulation model, which fully restores the real-world scene, making the decision-making actions of the experimenters infinitely close to the real situation. It reflects the different parking and location decision-making behaviors of users in different traffic environments at the micro level, which greatly facilitates system modeling and related research in the field of traffic road patterns. Moreover, the Unity simulation model can obtain real-time global information data. The data collection not only has a latency of only 60 milliseconds, but also the information acquisition is very complete and accurate.

[0036] Second, this invention outputs the vehicle's driving path in the form of a trajectory. This trajectory pattern is suitable for various subsequent algorithm processing, including but not limited to clustering algorithms, inverse reinforcement learning algorithms, and reinforcement learning algorithms, which helps to systematically model and study traffic road patterns.

[0037] Third, this invention is applicable to various traffic road environments, including surface traffic and three-dimensional traffic, and is beneficial for traffic road simulation and data collection in various situations. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the urban road network simulation model of the present invention;

[0039] Figure 2 This is a schematic diagram of the method flow of the present invention;

[0040] Figure 3 This is a schematic diagram of the modeling skeleton of the road network and urban buildings of the present invention;

[0041] Figure 4 This is a schematic diagram of a simulated first-person perspective of the present invention;

[0042] Figure 5 This is a schematic diagram of the data output of the present invention;

[0043] Figure 6 This is a schematic diagram illustrating the linear effect of the connection between the present invention and the inverse reinforcement learning algorithm;

[0044] Figure 7 This is a schematic diagram illustrating the graphical effect of the integration of the present invention with the inverse reinforcement learning algorithm. Detailed Implementation

[0045] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0046] Example

[0047] like Figure 1As shown, this invention uses game engine modeling and simulation software to construct a complete urban road network simulation model. This model has the following functions: a traffic simulation system that simulates real-time urban road network traffic flow and roadside parking systems; a driving simulation system that allows drivers to complete parking space search simulation experiments from a first-person perspective; and a data acquisition and analysis system that performs real-time filtering and acquisition of driver experimental trajectory data and global traffic information data, enabling subsequent data processing and model training using various algorithms. This invention can be further subdivided into an urban traffic system, a roadside parking system, a navigation map system, a driving simulation system, and a scene simulation system.

[0048] like Figure 2 As shown, the present invention provides a method for constructing an urban road network simulation model for parking space search behavior experiments. The method includes the following steps:

[0049] S1, such as Figure 3 As shown, a traffic simulation system was built using the Unity game engine, constructing a fully simulated scene model, including road surfaces, parking space markings, and basic building systems, to simulate urban road network traffic flow and roadside parking systems. The specific steps are as follows:

[0050] S101. Select the target road network and model it using the Unity game engine. Extract the road network information into a graph structure composed of points and edges, and store urban traffic rule information based on points and edges. Represent the road segments and intersections of the road network information as edges and points respectively, forming a large-scale urban road network with a graph structure. Store various types of road network information in the edges and points, such as road segment location, road segment orientation, current road segment traffic flow, current road segment parking space occupancy rate, etc.

[0051] S102. The simulated urban road network traffic flow is abstracted into specific vehicles. The vehicles are abstracted into a separate class _childRoads in the game engine script function. They are given vehicle information such as vehicle location information, vehicle speed information, and vehicle volume information. The relevant data information of the vehicles will be automatically matched with the road segment information where the vehicles are located in the platform, thereby forming a complete urban road network traffic flow through the dynamic interaction between vehicles and road network nodes.

[0052] S103. Taking road segments as the basic operation unit, and road segment feature information as traffic rule information and real-time traffic flow information of urban road network, a logical subject is established as the main traffic rule carrier, giving road segments an executable operation space.

[0053] S104. Assign pathfinding algorithms to vehicle units simulating urban road network traffic flow, including a global optimal path selection algorithm, a local optimal path selection algorithm, and a path selection algorithm based on specific population preferences, forming a diverse and complex urban road network traffic flow. The global optimal path selection algorithm is implemented using the Floyd-Warshall algorithm, employing an n*n matrix `map` to record the shortest path information in the road network. `map[i, j]` represents the shortest distance between road segment numbered `i` and road segment numbered `j`. This matrix information is updated every frame, and its state transition equation is as follows:

[0054] map[i,j]=min{map[i,k]+map[k,j]+map[i,j]}

[0055] The local optimal path selection algorithm uses a greedy algorithm:

[0056] temp1=min{_chilRoads[0].carnum, _chilRoads[1].carnum}

[0057] temp2=min{_chilRoads[1].carnum, _chilRoads[2].carnum}

[0058] temp1 = min{temp1, temp2}

[0059] temp1=min{_chilRoads[3].carnum, temp1}

[0060] The specific population preference-based path selection algorithm uses reinforcement learning to obtain the policy function and calculates the path selection parameters frame by frame.

[0061] S105. To enable vehicle units simulating urban road network traffic flow to detect obstacles around the vehicle, an arc-shaped detection surface is created based on the ray-detection mechanism of a game engine. The ray density, detection angle range, and number of ray layers in the control are adjusted through parameter settings, allowing vehicles to detect obstacles ahead, whether vehicles or buildings, ensuring that vehicles do not collide or cause congestion. The calculation method for the arc-shaped detection surface is as follows: lookAngle and lookAccurate control the ray density and detection angle range of the arc-shaped detection surface, and i is the number of ray layers in the cyclic control.

[0062]

[0063] ArrayA=Quaternion.Euler(0,-1*subAngle*(i+1),0)

[0064] ArrayB=Quaternion.Euler(0, subAngle*(i+1), 0).

[0065] S2, such as Figure 4 As shown, a driving simulation system is established to simulate the driving environment within the traffic simulation system and provide real-time road condition navigation. A parking space finding simulation experiment is completed from a first-person perspective. The specific steps are as follows:

[0066] S201. Construct a basic acceleration control framework to enable the driver to control the direction and speed of the vehicle via the keyboard. The W key imparts forward acceleration to the vehicle, the S key imparts acceleration opposite to the current direction of the vehicle's movement, the A key controls the vehicle to turn left, and the D key controls the vehicle to turn right.

[0067] S202. After repeated real-person experiments, the driver and vehicle driving-related variables, such as forward acceleration, were adjusted to make the vehicle driving closer to the real situation.

[0068] S203. Based on the existing physical entity simulation function of the game engine, realize the detection of collisions between the driver vehicle and objects in the traffic simulation system established in S1, as well as the handling of special boundary cases.

[0069] S204. The driver's vehicle is equipped with an automatic following function for the game engine navigation camera;

[0070] S205. Through layered visualization, the position of the vehicle driven by the experimenter is magnified on the map, and the traffic flow information on the map is updated in real time. The direction vector from the current position to the target is calculated by subtracting the target position vector DES.transform.position from the current driver's vehicle position vector BV.transform.position. The top right corner of the screen displays a top-down view of a circular area centered on the driver's vehicle, composed of the radius vector maplength. The lengths of maplength in the same direction as direction are compared, and the vector with the smaller length is taken as the indicator vector. The end of the vector is marked as the target location guide marker, with the driver's vehicle as the starting point. This achieves automatic target location guidance calculation.

[0071] direction=(DES.trasform.position-BV.transform.position).magnitude

[0072] distance=min{direction,maplength}

[0073] transform.position=Vector3*(distance.x, height, distance.z).

[0074] S3, such as Figure 5 As shown, a data acquisition and analysis system is established to collect and filter trajectory data and global traffic information data from the parking space search simulation experiment in real time, and convert them into an expert trajectory dataset. This expert trajectory dataset is then input into various algorithms for data processing and model training to obtain the policy function; for example... Figure 6 and 7 The diagram illustrates the effect of connecting the present invention with subsequent algorithms, using inverse reinforcement learning as an example. The subsequent algorithms include, but are not limited to, clustering algorithms, inverse reinforcement learning algorithms, and reinforcement learning algorithms.

[0075] S301. Obtain real-time global traffic information through urban road network traffic flow, perform data correctness detection based on traffic rules, and preprocess the data to form a data test set with trajectory as the basic data format.

[0076] S302. By matching the coordinates of all vehicles in the experiment with the coordinate range of road segments and parking spaces, the real-time traffic flow and parking information of each road segment are statistically obtained, thereby constructing global feature information.

[0077] S303. By matching the real-time coordinates of the vehicle controlled by the experimenter with the road segment coordinate range, the location information of the controlled vehicle on the road segment is obtained in real time, and the information is saved to construct an expert trajectory dataset.

[0078] S304. Input the obtained trajectory data into the algorithm to train the model and obtain the policy function.

[0079] S4. Solve for vehicle information and global traffic information using a strategy function to form a single action command that guides the vehicle's next movement strategy, specifically:

[0080] The entire strategy function is stored in a background file. When a vehicle in the traffic simulation stream arrives at the decision intersection, the current vehicle information and global road information are automatically input into the strategy function to solve for the solution. The solution forms a single action command, including a straight-ahead command, a left-turn command, a right-turn command, and a U-turn command. This action command guides the vehicle's next movement strategy.

[0081] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for constructing a simulation model of an urban road network for parking behavior experiments, characterized in that, The method includes the following steps: S1. Establish a traffic simulation system to simulate urban road network traffic flow and roadside parking system; S2. Establish a driving simulation system to simulate the driving environment in the traffic simulation system in S1 and provide real-time road condition navigation. Complete the parking space finding simulation experiment from a first-person perspective. S3. Establish a data acquisition and analysis system to collect and filter the trajectory data and global traffic information data of the parking space search simulation experiment in S2 in real time, and convert them into expert trajectory datasets. Input the expert trajectory datasets into various algorithms to process the data and train the models to obtain the policy functions. S4. Solve the vehicle information and global traffic information through the strategy function to form a single action command to guide the vehicle's next movement strategy; The specific steps of step S3 are as follows: S301. Obtain real-time global traffic information through urban road network traffic flow, perform data correctness detection based on traffic rules, and preprocess the data to form a data test set with trajectory as the basic data format. S302. By matching the coordinates of all vehicles in the experiment with the coordinate range of road segments and parking spaces, the real-time traffic flow and parking information of each road segment are statistically obtained, thereby constructing global feature information. S303. By matching the real-time coordinates of the vehicle controlled by the experimenter with the road segment coordinate range, the location information of the controlled vehicle on the road segment is obtained in real time, and the information is saved to construct an expert trajectory dataset. S304. Input the obtained trajectory data into various algorithms to train the model and obtain the policy function; The various algorithms mentioned include clustering algorithms, inverse reinforcement learning algorithms, and reinforcement learning algorithms.

2. The method for constructing an urban road network simulation model for parking behavior experiments according to claim 1, characterized in that, The specific steps of S1 are as follows: S101. Select the target road network, use the game engine to model it, extract the road network information into a graph structure composed of points and edges, and store urban traffic rule information with points and edges as the basic units. S102. Abstract the traffic flow of the simulated city road network into specific vehicles, and abstract the vehicles into separate classes in the game engine script functions. The system assigns vehicle information, which is then matched with the road segment information where the vehicle is located. Through the dynamic interaction between the vehicle and the road network nodes, a complete urban road network traffic flow is established. The vehicle information includes vehicle location information, vehicle speed information, and vehicle volume information. S103. Taking road segments as the basic operation unit, and road segment feature information as traffic rule information and real-time traffic flow information of urban road network, a logical subject is established as the main traffic rule carrier, giving road segments an executable operation space. S104. Assign a pathfinding algorithm to vehicle units that simulate urban road network traffic flow to form diverse and complex urban road network traffic flow. S105. To enable vehicle units that simulate urban road network traffic flow to detect obstacles around the vehicle, an arc-shaped detection surface is created based on the ray detection mechanism of the game engine. The ray density, detection angle range, and number of ray layers in the control are controlled by parameter settings.

3. The method for constructing an urban road network simulation model for parking behavior experiments according to claim 2, characterized in that, The pathfinding algorithms in step S104 include a global optimal path selection algorithm, a local optimal path selection algorithm, and a path selection algorithm based on specific population preferences. The path selection algorithm based on specific population preferences uses reinforcement learning to obtain a policy function and calculates the path selection parameters frame by frame.

4. The method for constructing an urban road network simulation model for parking behavior experiments according to claim 1, characterized in that, The specific steps of S2 are as follows: S201. Construct a basic acceleration control framework to enable the driver to control the vehicle's direction and speed via keyboard operation. S202. After repeated real-person experiments and adjustments to the driver and vehicle driving-related variables, the vehicle driving is made closer to the real situation. S203. Based on the existing physical entity simulation function of the game engine, realize the detection of collisions between the driver vehicle and objects in the traffic simulation system established in S1, as well as the handling of special boundary cases. S204. The driver's vehicle is equipped with an automatic following function for the game engine navigation camera; S205. By using layered visualization, the location of the vehicle driven by the experimenter is magnified on the map, and the traffic flow information on the map is updated in real time to achieve automatic guidance and calculation of the target location.

5. The method for constructing an urban road network simulation model for parking behavior experiments according to claim 4, characterized in that, The keyboard operations in step S201 are as follows: the W key imparts forward acceleration to the vehicle, the S key imparts acceleration to the vehicle in the opposite direction to the current direction of the vehicle's direction, the A key controls the vehicle to turn left, and the D key controls the vehicle to turn right.

6. The method for constructing an urban road network simulation model for parking behavior experiments according to claim 1, characterized in that, Step S4 specifically involves storing the entire strategy function in a background file. When a vehicle in the traffic simulation stream reaches the decision intersection, the current vehicle information and global traffic information are automatically input into the strategy function for solving. The resulting solution forms a single action command to guide the vehicle's next movement strategy.

7. The method for constructing an urban road network simulation model for parking behavior experiments according to claim 6, characterized in that, The action commands include straight-ahead command, left-turn command, right-turn command, and U-turn command.

Images