Traffic optimization evaluation method and system based on radar and vision integrated machine and AI algorithm

Abstract

The application provides a traffic optimization evaluation method and system based on a radar and vision integrated machine and an AI algorithm, which comprises the following steps: performing data preprocessing on multi-source heterogeneous sensing data to generate a high-precision fusion trajectory set; constructing a SUMO simulation environment according to a road network file, a traffic flow file and a simulation configuration file; importing the high-precision fusion trajectory set into the SUMO simulation environment; at each control decision time, reading and constructing a state vector from the radar and vision integrated machine or the SUMO simulation environment, selecting a signal control scheme according to the traffic demand of different scenes and time periods, and generating a signal control strategy; constructing a set of schemes to be evaluated, and for each scheme to be evaluated, running the SUMO simulation under the same road network, the same traffic demand and the same evaluation interval, collecting simulation output results, calculating traffic efficiency indicators, traffic safety and emission evaluation indicators, and obtaining comparative analysis results among different schemes. The application provides more complete traffic flow data than single vision or radar.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a traffic optimization evaluation method and system based on a radar-visual integrated machine and AI algorithms. Background Technology

[0002] As urban traffic complexity increases, traffic management departments are becoming increasingly reliant on simulation assessments and decision optimization. Traditional traffic simulation assessment methods often depend on fixed detectors or manually labeled data, making it difficult to accurately reflect dynamic road changes, and there are significant discrepancies between simulation model parameters and real traffic behavior.

[0003] The LiDAR-Camera Fusion Sensor integrates LiDAR and visual images to achieve high-precision target recognition, vehicle trajectory extraction, and traffic event detection, providing refined measured input for traffic simulation models. However, due to limitations such as sensor installation location, weather conditions, and scene complexity, its data may still deviate from the virtual traffic flow generated by SUMO simulation.

[0004] Existing research mostly focuses on calibration of a single data source or optimization of simulation parameters, lacking a comprehensive evaluation system that can form a closed-loop adaptive optimization among the perception layer, algorithm layer and simulation layer.

[0005] In existing technologies, single-sensor perception is limited by its own characteristics, making it difficult to achieve high-precision, full-dimensional identification of traffic elements in complex scenarios; multi-source sensor fusion often remains at a shallow level of data registration, lacking in-depth mining of the spatiotemporal correlation of cross-modal data, resulting in insufficient robustness and completeness of the fusion results; SUMO traffic simulation platforms mostly rely on static preset parameters or historical statistical data, with weak dynamic perception data access capabilities, and the simulation modeling of individual traffic participants' behaviors deviates significantly from reality; at the same time, there is a technological gap between traffic behavior modeling and simulation evaluation, making it difficult to form a closed-loop link of "perception, modeling, and simulation," thus limiting the authenticity and application value of simulation evaluation results. Summary of the Invention

[0006] Based on this, the purpose of the present invention is to provide a traffic optimization evaluation method and system based on a radar-view integrated machine and AI algorithm, so as to at least solve the shortcomings of the above-mentioned technology.

[0007] This invention proposes a traffic optimization evaluation method based on a radar-visual integrated machine and AI algorithms, comprising: The system acquires LiDAR point cloud and camera image data through the integrated laser vision device, and extracts vehicle position, speed, acceleration and trajectory information through a spatiotemporal synchronization mechanism to obtain dynamic traffic flow data on the road in real time, so as to form corresponding multi-source heterogeneous perception data. The multi-source heterogeneous sensing data is preprocessed to generate a high-precision fused trajectory set; Generate a road network file, a traffic flow file, and a simulation configuration file, and construct a SUMO simulation environment based on the road network file, the traffic flow file, and the simulation configuration file; The high-precision fused trajectory set is imported into the SUMO simulation environment. At the start of the simulation, vehicles are generated according to their departure time, route, and vehicle type parameters and injected into the SUMO simulation environment. The current traffic status data is output and traffic control commands are received through the TraCI server interface. The simulation status and result data are continuously recorded and output. At each control decision, a state vector is read and constructed from the radar-visual integrated machine or the SUMO simulation environment, and a traffic control scheme is selected according to the traffic demand of different scenarios and time periods to generate a traffic control strategy. A set of schemes to be evaluated is constructed. For each scheme, SUMO simulation is run under the same road network, the same traffic demand, and the same evaluation interval. The simulation output results are collected, and traffic efficiency indicators, traffic safety and emission evaluation indicators are calculated. The indicator results of each scheme are normalized and statistically summarized to obtain comparative analysis results between different schemes.

[0008] Furthermore, the step of preprocessing the multi-source heterogeneous sensing data to generate a high-precision fused trajectory set includes: The target sequence in the multi-source heterogeneous sensing data is filtered by a combination of rules and statistics, and the trajectory breaks caused by short-term frame loss or occlusion in the multi-source heterogeneous sensing data are completed by a piecewise linear interpolation method. Given a set of sample points sorted in ascending order of the independent variable:

[0009] interval Then the piecewise linear interpolation is:

[0010] In the formula, This represents the value of the independent variable to be interpolated. This indicates that when the independent variable is Interpolated estimates at time; The video detection results in the multi-source heterogeneous sensing data are transformed from pixel coordinates to world coordinates, and the radar targets are transformed from polar coordinates to world coordinates, and unified into state variables under the same coordinate system. At the same time, radar tracks and video tracks are correlated to obtain a set of matching pairs. For the matching targets, extended Kalman filtering is used for fusion and updating to obtain a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained and the source and confidence level are marked. For ID switching or broken trajectories, trajectory splicing and reconstruction are performed based on spatiotemporal continuity, lane constraints, and speed trend consistency to form the final trajectory set.

[0011] Furthermore, the steps of generating road network files, traffic flow files, and simulation configuration files, and constructing a SUMO simulation environment based on the road network files, traffic flow files, and simulation configuration files include: Maps are generated by downloading maps from open-source atlases and then converted into road network files using the netconvert program included in the SUMO simulation. A flow rate file is generated using the multi-source heterogeneous sensing data, and a corresponding simulation configuration file is also generated. A SUMO simulation environment is constructed based on the road network file, the traffic flow file, and the simulation configuration file. Detectors are set in the simulation environment according to the actual deployment location of the radar-visual integrated machine, and the TraCI server is started.

[0012] Furthermore, the signal control scheme includes a system based on... Signal control schemes and adaptive control algorithm schemes for learning algorithms; Wherein, the basis The signal control scheme for the learning algorithm is as follows: Based on state Using the greedy method to select actions And perform the action to receive a reward. The environment has entered a new state. , choose to The largest As To update the value function, the state-action value is updated using the Bellman equation:

[0013] in, Indicates the learning rate. Indicates the discount factor; The adaptive control algorithm scheme is as follows: During off-peak hours or other periods with fewer vehicles, an adaptive algorithm is used to detect the number of vehicles entering each lane through the integrated radar-visual system and automatically extend the green light time.

[0014] Furthermore, the traffic efficiency indicators include at least average vehicle delay, average travel time, queue length, throughput, average vehicle speed, and number of stops.

[0015] This invention also proposes a traffic optimization evaluation system based on a radar-visual integrated machine and AI algorithms, comprising: The data acquisition module is used to acquire lidar point cloud and camera image data through the integrated laser vision device, and extract vehicle position, speed, acceleration and trajectory information through a spatiotemporal synchronization mechanism to acquire dynamic traffic flow data on the road in real time, so as to form corresponding multi-source heterogeneous perception data. The data processing module is used to preprocess the multi-source heterogeneous sensing data to generate a high-precision fusion trajectory set; The data simulation module is used to generate road network files, traffic flow files, and simulation configuration files, and to construct a SUMO simulation environment based on the road network files, traffic flow files, and simulation configuration files. The environment simulation module is used to import the high-precision fused trajectory set into the SUMO simulation environment. At the start of the simulation, it generates vehicles according to the vehicle departure time, route, and vehicle type parameters and injects them into the SUMO simulation environment. It outputs the current traffic status data and receives traffic control commands through the TraCI server interface, and continuously records and outputs the simulation status and result data. The strategy generation module is used to read and construct a state vector from the radar-visual integrated machine or the SUMO simulation environment at each control decision, select a traffic control scheme according to the traffic demand of different scenarios and time periods, and generate a traffic control strategy. The comparative analysis module is used to construct a set of schemes to be evaluated. For each scheme, SUMO simulation is run under the same road network, the same traffic demand, and the same evaluation interval. The simulation output results are collected, and traffic efficiency indicators, traffic safety and emission evaluation indicators are calculated. The indicator results of each scheme are normalized and statistically summarized to obtain comparative analysis results between different schemes.

[0016] Furthermore, the data processing module is specifically used for: The target sequence in the multi-source heterogeneous sensing data is filtered by a combination of rules and statistics, and the trajectory breaks caused by short-term frame loss or occlusion in the multi-source heterogeneous sensing data are completed by a piecewise linear interpolation method. Given a set of sample points sorted in ascending order of the independent variable:

[0017] interval Then the piecewise linear interpolation is:

[0018] In the formula, This represents the value of the independent variable to be interpolated. This indicates that when the independent variable is Interpolated estimates at time; The video detection results in the multi-source heterogeneous sensing data are transformed from pixel coordinates to world coordinates, and the radar targets are transformed from polar coordinates to world coordinates, and unified into state variables under the same coordinate system. At the same time, radar tracks and video tracks are correlated to obtain a set of matching pairs. For the matching targets, extended Kalman filtering is used for fusion and updating to obtain a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained and the source and confidence level are marked. For ID switching or broken trajectories, trajectory splicing and reconstruction are performed based on spatiotemporal continuity, lane constraints, and speed trend consistency to form the final trajectory set.

[0019] Furthermore, the data simulation module is specifically used for: Maps are generated by downloading maps from open-source atlases and then converted into road network files using the netconvert program included in the SUMO simulation. A flow rate file is generated using the multi-source heterogeneous sensing data, and a corresponding simulation configuration file is also generated. A SUMO simulation environment is constructed based on the road network file, the traffic flow file, and the simulation configuration file. Detectors are set in the simulation environment according to the actual deployment location of the radar-visual integrated machine, and the TraCI server is started.

[0020] The present invention also proposes a storage medium on which a computer program is stored, which, when executed by a processor, implements the above-mentioned traffic optimization evaluation method based on the radar-visual integrated machine and AI algorithm.

[0021] The present invention also proposes a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned traffic optimization evaluation method based on the radar-visual integrated machine and AI algorithm.

[0022] Compared with existing technologies, the traffic optimization evaluation method and system based on radar-visual integrated machine and AI algorithm in this invention achieves point cloud and image fusion based on radar-visual integrated machine, providing more complete traffic flow data than single vision or radar. It integrates the geometric information of radar and the semantic features of vision, and achieves deep coupling between the two through a spatiotemporal consistency fusion mechanism. Through precise coordinate calibration and feature weighting, it fully leverages the complementarity of data, which can overcome the environmental sensitivity of visual sensors and make up for the lack of semantic information of radar. This makes the identification of the location, speed, category and behavioral features of traffic elements more comprehensive and stable, providing high-quality basic data for simulation evaluation. It uses AI algorithm to realize dynamic deviation correction between simulation and real perception, improving the credibility of simulation data. Combining genetic algorithm and deep learning framework, the SUMO simulation system has self-learning and cross-scene transfer capabilities. It provides a unified verification environment for AI algorithms such as traffic signal optimization and intersection decision control, improving the repeatability and versatility of experiments. Attached Figure Description

[0023] Figure 1 This is a flowchart of the traffic optimization evaluation method based on the radar-visual integrated machine and AI algorithm in the first embodiment of the present invention; Figure 2 This is a structural block diagram of the traffic optimization and evaluation system based on the radar-visual integrated machine and AI algorithm in the second embodiment of the present invention; Figure 3 This is a structural block diagram of the computer in the third embodiment of the present invention.

[0024] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0025] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0027] Example 1 Please see Figure 1The diagram illustrates a traffic optimization evaluation method based on a radar-visual integrated machine and AI algorithms in the first embodiment of the present invention. The method specifically includes steps S101 to S106: S101 acquires LiDAR point cloud and camera image data through the integrated laser vision device, and extracts vehicle position, speed, acceleration and trajectory information through a spatiotemporal synchronization mechanism to obtain dynamic traffic flow data on the road in real time, so as to form corresponding multi-source heterogeneous perception data. In practice, initialization and parameter configuration are performed first, specifically: (1) Deployment information configuration: Configure basic information of intersections and road segments, including but not limited to: intersection center point coordinates, number of approach lanes, number of lanes, lane width, speed limit, stop line position, phase set, yellow light / all red light duration, minimum green light / maximum green light, control cycle, etc.

[0028] (2) Loading equipment calibration parameters: Number and associate each radar-visual integrated unit with each entrance lane, and establish a database to store the data indicators of each lane collected by each radar-visual integrated unit. Set the working parameters of each radar-visual integrated unit, load the internal and external parameter calibration results of radar and camera, and establish a unified coordinate system with the stop line as the origin.

[0029] (3) Time synchronization mechanism enabled: Start a unified clock source (which can be PTP / NTP / hardware PPS clock) and define a unified timestamp format. (milliseconds or microseconds) serve as the benchmark for subsequent fusion, simulation injection, and control closed-loop.

[0030] (4) Generate experimental configuration file: Generate the configuration file (such as JSON / YAML) for this simulation evaluation, including: data file path, simulation step size Control decision-making step size Evaluation intervals, random seeds, and lists of comparison schemes are used to ensure that the results are reproducible.

[0031] Furthermore, the radar-visual integrated machine simultaneously acquires millimeter-wave radar spot patterns, target lists, and video image frames to form a raw data stream. The radar data includes a target list. This includes distance, azimuth, radial velocity, confidence level, etc. Video data includes image frames. .

[0032] (2) Traffic target detection and tracking On the video side, target detection and multi-target tracking are performed on I(t) to obtain... The fields include category, 2D / 3D bounding box, pixel trajectory, confidence level, etc.; the radar side performs target tracking on R(t) to obtain... .

[0033] (3) Spatiotemporal synchronization and timestamp management Add a uniform timestamp to the data from each sensor. If frame rate differences exist, a "nearest neighbor time alignment + interpolation" strategy is used to align the radar time to the video frame time (or vice versa) to ensure consistent fusion time.

[0034] (4) Real-time collection of traffic flow parameters Calculate or output key traffic flow parameters at the data acquisition point or edge. Data such as traffic flow, occupancy rate, vehicle speed distribution, time headway, and queue length are output synchronously with target-level data.

[0035] S102, perform data preprocessing on the multi-source heterogeneous sensing data to generate a high-precision fusion trajectory set; Furthermore, step S102 specifically includes steps S1021 to S1024: S1021, the target sequence in the multi-source heterogeneous sensing data is filtered by a combination of rules and statistics, and the trajectory breaks caused by short-term frame loss or occlusion in the multi-source heterogeneous sensing data are completed by a piecewise linear interpolation method. S1022, Given a set of sample points sorted in ascending order of the independent variable:

[0036] interval Then the piecewise linear interpolation is:

[0037] In the formula, This represents the value of the independent variable to be interpolated. This indicates that when the independent variable is Interpolated estimates at time; S1023, the video detection results in the multi-source heterogeneous sensing data are transformed from pixel coordinates to world coordinates, and the radar target is transformed from polar coordinates to world coordinates, and unified into state variables under the same coordinate system; S1024: At the same time, radar tracks and video tracks are correlated to obtain a set of matching pairs. For the matching targets, extended Kalman filtering is used for fusion and updating to obtain a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained and the source and confidence level are marked. For ID switching or broken trajectories, trajectory splicing and reconstruction are performed based on spatiotemporal continuity, lane constraints, and speed trend consistency to form the final trajectory set.

[0038] In practice, noise and anomaly removal are performed on the data. Target sequences are filtered using a combination of rules and statistics. This includes: 1. Applying physical constraints on velocity / acceleration; 2. Removing targets with excessively short trajectory lengths, large jumps, or duplicate IDs; 3. Removing radar false alarms and video misdetections based on confidence thresholds and temporal consistency. Furthermore, for trajectory breaks caused by short-term frame drops or occlusion, spatiotemporal interpolation is used to complete the trajectory and generate continuous trajectory segments.

[0039] The video detection results are transformed from pixel coordinates to world coordinates (using calibration matrix / homography transformation), and the radar target is transformed from polar coordinates to world coordinates, and then unified into state variables under the same coordinate system.

[0040] At the same time Data association is performed between radar tracks and video tracks to obtain a set of matching pairs. An extended Kalman filter is used to fuse and update the matched target, resulting in a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained but their source and confidence level are labeled. For ID switching or broken trajectories, a trajectory splicing strategy based on spatiotemporal continuity, lane constraints, and speed trend consistency is used for reconstruction to form the final trajectory set. .

[0041] S103, Generate road network file, traffic flow file and simulation configuration file, and construct SUMO simulation environment based on the road network file, traffic flow file and simulation configuration file; Furthermore, step S103 specifically includes steps S1031 to S1033: S1031 generates a map file by downloading it from an open-source atlas and converts it into a road network file using the netconvert program included in the SUMO simulation. S1032, using the multi-source heterogeneous sensing data, a flow file is generated, and a corresponding simulation configuration file is generated. S1033, construct a SUMO simulation environment based on the road network file, the traffic file and the simulation configuration file, set detectors in the simulation environment according to the actual deployment location of the radar-visual integrated machine, and start the TraCI server.

[0042] In practical implementation, a SUMO simulation is constructed, specifically: (1) Simulation environment setup SUMO simulation requires three basic files: a road network file (.net.xml), a traffic flow file (.rou.xml), and a simulation configuration file (.sumocfg). The road network file defines the road infrastructure, including road network geometry, topological connections, and traffic rules. The traffic flow file defines traffic demand and vehicle trips, including vehicle type, trip demand, driving behavior, and origin-destination (OD) parameters. The simulation configuration file is used to set simulation time settings, simulation behavior, and output settings. There are three methods for constructing road network files. First, you can download a map generation (.osm) file from the open-source map atlas OpenStreetMap and then convert it into a road network file (.net.xml) using the netconvert program included with Sumo. Second, you can manually label key information such as lanes, intersections, and traffic lights in the netconvert visualization interface. Third, you can directly define edges, nodes, connections, signals, and timings in the file.

[0043] The traffic flow file is generated from data collected by the radar-visual integrated machine. The simulation configuration file sets up the detectors based on the actual deployment of the radar-visual integrated machine.

[0044] (2) Detector settings Detector settings can be configured via an additional file (.add.xml), which is then loaded by the simulation configuration file (.sumocfg). The detector settings are configured according to the actual deployment location of the radar-visual integrated machine, and simulation data is output in real time during the simulation process.

[0045] (3) Simulation scheduling settings Set simulation step size Simulation duration Output granularity (per step / per second / per cycle) and enable the TraCI server for external algorithm connection and control.

[0046] S104, import the high-precision fused trajectory set into the SUMO simulation environment. At the start of the simulation, generate vehicles according to vehicle departure time, route, and vehicle type parameters and inject them into the SUMO simulation environment. Output the current traffic status data and receive traffic control instructions through the TraCI server interface. Continuously record and output simulation status and result data. In practice, the generated traffic flow input file is imported into the SUMO simulation module. At the start of the simulation, vehicles are generated and injected into the simulation environment according to information such as vehicle departure time, route, and vehicle type parameters. When the trajectory-driven method is used, vehicles are created at the corresponding time according to a unified timestamp and the vehicle status is updated based on trajectory position and speed parameters, thereby reproducing the real traffic operation characteristics of the target time period.

[0047] During the simulation, the simulation process scheduling and control unit is responsible for advancing the simulation time and the evolution of vehicle behavior. It advances the simulation clock according to the preset simulation step size and outputs the current traffic status data to the signal control algorithm module through the TraCI interface in each control cycle. At the same time, it receives signal control commands such as phase switching and green light duration adjustment issued by the signal control algorithm module and writes them into the traffic light controller to realize the closed-loop linkage between simulation and control.

[0048] Throughout the entire simulation cycle, the system continuously records and outputs status and result data such as traffic flow, delays, queue lengths, average travel time, number of stops, and emissions, providing basic data for subsequent simulation evaluation module index calculations and scheme comparison analysis.

[0049] S105, at each control decision, read and construct a state vector from the radar-visual integrated machine or the SUMO simulation environment, select a traffic control scheme according to the traffic demand of different scenarios and time periods, and generate a traffic control strategy; In practice, at each control decision moment Read and construct state vectors from the Rave-Video all-in-one machine or SUMO simulation. This includes: queue length, waiting time, occupancy rate, lane arrival rate, saturation, phase remaining time, average vehicle speed, and number of stops for each entrance lane / lane.

[0050] Different signal control solutions can be selected based on usage requirements for different scenarios and time periods.

[0051] (1) Signal control scheme based on Q-learning algorithm Q-learning is a method that uses temporal difference to solve reinforcement learning control problems. Given the five elements of reinforcement learning, it uses two control policies to update the selected new actions and value functions in order to solve for the optimal state-action value function and the optimal policy.

[0052] Q-learning algorithm is state-based. Using the greedy method to select actions And perform the action to receive a reward. At this point, the environment enters a new state. , choose to The largest As The value function is updated using the recursive nature of the Bellman equation and the Markov property. The value, whose expression is:

[0053] in, Indicates the learning rate. Indicates the discount factor; The algorithm structure is as follows: 1. Randomly initialize all parameters of the Q network; 2. For episode = 1 to M; 3. The agent obtains the initial environmental state S0; 4. For t=0 to T; 5. According to Strategy, based on probability Select a random action from the action space; 6. Perform the current action while in this state to receive a reward. and new status ; 7. Store experience samples in the experience pool. , , , ); 8. Use value networks for prediction, inputting environmental conditions. Output the Q value of the action performed. ; 9. Use value network to calculate TD target, input state Output ; 10. Calculate the time difference error. ; 11. Calculate the loss function The gradient is then expressed as: ; 12. If the round reaches its maximum value and the state is terminated, the training ends; otherwise, training continues.

[0054] Furthermore, the adaptive control algorithm scheme is as follows: During off-peak hours or other periods with fewer vehicles, an adaptive algorithm is used to detect the number of vehicles entering each lane through the integrated radar-visual system and automatically extend the green light time.

[0055] Furthermore, the traffic efficiency indicators include at least average vehicle delay, average travel time, queue length, throughput, average vehicle speed, and number of stops. The algorithm structure is shown below: 1. Initialize the signal controller parameters: Set off-peak time window ; 2. For episode = 1 to M; 3. The agent obtains the initial environmental state S0; 4. For t=0 to T; 5. The number of vehicles entering each lane is obtained by the radar-guided integrated machine. And obtain the current phase With Green Time ; 6. Calculate the low demand indicator ; 7. If ,but else ; 8. If Proceed to step 14; 9. else calculates the number of oncoming vehicles that can be extended from the current phase. ,in Indicates the entrance lane that is currently open for passage; 10. If Maintain the current phase and continue to allow passage. Proceed to step 14; 11. If and Execute the green light extension; 12. else calculate other phase requirements ; 13. If End the current phase and switch to the phase with the highest demand. Simultaneously, the transition between the yellow and red light phases is implemented. 14. end for.

[0056] S106. Construct a set of schemes to be evaluated. For each scheme to be evaluated, run SUMO simulation under the same road network, the same traffic demand, and the same evaluation interval. Collect simulation output results, calculate traffic efficiency indicators, traffic safety and emission evaluation indicators, normalize and statistically summarize the indicator results of each scheme to be evaluated, and obtain comparative analysis results between different schemes.

[0057] In practical implementation, simulation evaluation and comparison of multiple solutions are conducted, specifically: (1) Simulation scheme settings In the simulation evaluation module, a set of schemes to be evaluated is constructed. The set of schemes includes baseline signal control schemes and artificial intelligence signal control schemes, and may further include perception input schemes corresponding to different numbers of radar screens, deployment locations or coverage areas, as well as different model parameter configuration schemes.

[0058] (2) Calculation of simulation indexes For each scheme, under the same road network, the same traffic demand, and the same evaluation interval, the SUMO simulation output results are collected, and the vehicle trajectory and statistical results data are calculated to calculate traffic efficiency indicators. The traffic efficiency indicators include at least average vehicle delay, average travel time, queue length, throughput, average vehicle speed, and number of stops.

[0059] (3) Fitting of calculation results A comprehensive assessment of traffic safety and emissions impacts is conducted based on vehicle speed and acceleration sequences, number of stops and starts, and conflict substitution indicators. The output includes extended indicators such as fuel consumption, emissions, or conflict frequency. The indicator results of each scheme are normalized and statistically summarized to obtain the differences in indicators, improvement ratios, and applicable conditions between different schemes. Comparative analysis results are formed to support the conclusions, thus providing a basis for the generation of assessment reports and the verification of the effectiveness of the methods.

[0060] In summary, the traffic optimization and evaluation method based on radar-visual integrated machine and AI algorithm in the above embodiments of the present invention provides more complete traffic flow data than single vision or radar by achieving point cloud and image fusion based on radar-visual integrated machine; it integrates the geometric information of radar and the semantic features of vision, and achieves deep coupling between the two through a spatiotemporal consistency fusion mechanism; it fully leverages the complementarity of data through precise coordinate calibration and feature weighting, which can overcome the environmental sensitivity of visual sensors and make up for the lack of semantic information of radar, making the identification of the position, speed, category and behavioral features of traffic elements more comprehensive and stable, providing high-quality basic data for simulation evaluation; it uses AI algorithm to realize dynamic deviation correction between simulation and real perception, improving the credibility of simulation data; it combines genetic algorithm and deep learning framework to enable SUMO simulation system to have self-learning and cross-scene transfer capabilities; and it provides a unified verification environment for AI algorithms such as traffic signal optimization and intersection decision control, improving the repeatability and versatility of experiments.

[0061] Example 2 Another aspect of this invention proposes a traffic optimization and evaluation system based on a radar-visual integrated machine and AI algorithms. Please refer to [link / reference needed]. Figure 2 The image shows a traffic optimization and evaluation system based on a radar-visual integrated machine and AI algorithm according to a second embodiment of the present invention. The system includes: The data acquisition module 11 is used to acquire lidar point cloud and camera image data through the laser vision all-in-one machine, and extract vehicle position, speed, acceleration and trajectory information through a spatiotemporal synchronization mechanism, and acquire dynamic traffic flow data on the road in real time to form corresponding multi-source heterogeneous perception data. Data processing module 12 is used to preprocess the multi-source heterogeneous sensing data to generate a high-precision fusion trajectory set; The data simulation module 13 is used to generate road network files, traffic flow files and simulation configuration files, and to construct a SUMO simulation environment based on the road network files, traffic flow files and simulation configuration files; The environment simulation module 14 is used to import the high-precision fused trajectory set into the SUMO simulation environment. At the start of the simulation, it generates vehicles according to the vehicle departure time, route, and vehicle type parameters and injects them into the SUMO simulation environment. It outputs the current traffic status data and receives traffic control instructions through the TraCI server interface, and continuously records and outputs the simulation status and result data. The strategy generation module 15 is used to read and construct a state vector from the radar-visual integrated machine or the SUMO simulation environment at each control decision, select a traffic control scheme according to the traffic demand of different scenarios and time periods, and generate a traffic control strategy. The comparative analysis module 16 is used to construct a set of schemes to be evaluated. For each scheme to be evaluated, SUMO simulation is run under the same road network, the same traffic demand and the same evaluation interval. The simulation output results are collected, and traffic efficiency index, traffic safety and emission evaluation index are calculated. The index results of each scheme to be evaluated are normalized and statistically summarized to obtain the comparative analysis results between different schemes.

[0062] Furthermore, the data processing module 12 is specifically used for: The target sequence in the multi-source heterogeneous sensing data is filtered by a combination of rules and statistics, and the trajectory breaks caused by short-term frame loss or occlusion in the multi-source heterogeneous sensing data are completed by a piecewise linear interpolation method. Given a set of sample points sorted in ascending order of the independent variable:

[0063] interval Then the piecewise linear interpolation is:

[0064] In the formula, This represents the value of the independent variable to be interpolated. This indicates that when the independent variable is Interpolated estimates at time; The video detection results in the multi-source heterogeneous sensing data are transformed from pixel coordinates to world coordinates, and the radar targets are transformed from polar coordinates to world coordinates, and unified into state variables under the same coordinate system. At the same time, radar tracks and video tracks are correlated to obtain a set of matching pairs. For the matching targets, extended Kalman filtering is used for fusion and updating to obtain a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained and the source and confidence level are marked. For ID switching or broken trajectories, trajectory splicing and reconstruction are performed based on spatiotemporal continuity, lane constraints, and speed trend consistency to form the final trajectory set.

[0065] Furthermore, the data simulation module 14 is specifically used for: Maps are generated by downloading maps from open-source atlases and then converted into road network files using the netconvert program included in the SUMO simulation. A flow rate file is generated using the multi-source heterogeneous sensing data, and a corresponding simulation configuration file is also generated. A SUMO simulation environment is constructed based on the road network file, the traffic flow file, and the simulation configuration file. Detectors are set in the simulation environment according to the actual deployment location of the radar-visual integrated machine, and the TraCI server is started.

[0066] The functions or operation steps implemented by the above modules and units are largely the same as those in the above method embodiments, and will not be repeated here.

[0067] The traffic optimization and evaluation system based on the radar-visual integrated machine and AI algorithm provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0068] Example 3 This invention also proposes a computer, please refer to [link / reference]. Figure 3 The computer shown in the third embodiment of the present invention includes a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the computer program 30, it implements the above-mentioned traffic optimization evaluation method based on the radar-visual all-in-one machine and AI algorithm.

[0069] The memory 10 includes at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10 can be an internal storage unit of a computer, such as the computer's hard disk. In other embodiments, the memory 10 can be an external storage device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Furthermore, the memory 10 can include both internal and external storage units of the computer. The memory 10 can be used not only to store application software and various types of data installed on the computer, but also to temporarily store data that has been output or will be output.

[0070] In some embodiments, the processor 20 may be an electronic control unit (ECU, also known as a vehicle computer), a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip, used to run program code stored in the memory 10 or process data, such as executing access restriction programs.

[0071] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the computer. In other embodiments, the computer may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0072] This invention also proposes a storage medium storing a computer program that, when executed by a processor, implements the traffic optimization evaluation method based on the radar-visual integrated machine and AI algorithm as described above.

[0073] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0074] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0075] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0076] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0077] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A traffic optimization evaluation method based on a radar-visual integrated machine and AI algorithms, characterized in that, include: The system acquires LiDAR point cloud and camera image data through the integrated laser vision device, and extracts vehicle position, speed, acceleration and trajectory information through a spatiotemporal synchronization mechanism to obtain dynamic traffic flow data on the road in real time, so as to form corresponding multi-source heterogeneous perception data. The multi-source heterogeneous sensing data is preprocessed to generate a high-precision fused trajectory set; Generate a road network file, a traffic flow file, and a simulation configuration file, and construct a SUMO simulation environment based on the road network file, the traffic flow file, and the simulation configuration file; The high-precision fused trajectory set is imported into the SUMO simulation environment. At the start of the simulation, vehicles are generated according to their departure time, route, and vehicle type parameters and injected into the SUMO simulation environment. The current traffic status data is output and traffic control commands are received through the TraCI server interface. The simulation status and result data are continuously recorded and output. At each control decision, a state vector is read and constructed from the radar-visual integrated machine or the SUMO simulation environment, and a traffic control scheme is selected according to the traffic demand of different scenarios and time periods to generate a traffic control strategy. A set of schemes to be evaluated is constructed. For each scheme, SUMO simulation is run under the same road network, the same traffic demand, and the same evaluation interval. The simulation output results are collected, and traffic efficiency indicators, traffic safety and emission evaluation indicators are calculated. The indicator results of each scheme are normalized and statistically summarized to obtain comparative analysis results between different schemes.

2. The traffic optimization evaluation method based on radar-visual integrated machine and AI algorithm according to claim 1, characterized in that, The steps of preprocessing the multi-source heterogeneous sensing data to generate a high-precision fused trajectory set include: The target sequence in the multi-source heterogeneous sensing data is filtered by a combination of rules and statistics, and the trajectory breaks caused by short-term frame loss or occlusion in the multi-source heterogeneous sensing data are completed by a piecewise linear interpolation method. Given a set of sample points sorted in ascending order of the independent variable: interval Then the piecewise linear interpolation is: In the formula, This represents the value of the independent variable to be interpolated. This indicates that when the independent variable is Interpolated estimates at time; The video detection results in the multi-source heterogeneous sensing data are transformed from pixel coordinates to world coordinates, and the radar targets are transformed from polar coordinates to world coordinates, and unified into state variables under the same coordinate system. At the same time, radar tracks and video tracks are correlated to obtain a set of matching pairs. For the matching targets, extended Kalman filtering is used for fusion and updating to obtain a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained and the source and confidence level are marked. For ID switching or broken trajectories, trajectory splicing and reconstruction are performed based on spatiotemporal continuity, lane constraints, and speed trend consistency to form the final trajectory set.

3. The traffic optimization evaluation method based on radar-visual integrated machine and AI algorithm according to claim 1, characterized in that, The steps of generating a road network file, a traffic flow file, and a simulation configuration file, and constructing a SUMO simulation environment based on the road network file, the traffic flow file, and the simulation configuration file include: Maps are generated by downloading maps from open-source atlases and then converted into road network files using the netconvert program included in the SUMO simulation. A flow rate file is generated using the multi-source heterogeneous sensing data, and a corresponding simulation configuration file is also generated. A SUMO simulation environment is constructed based on the road network file, the traffic flow file, and the simulation configuration file. Detectors are set in the simulation environment according to the actual deployment location of the radar-visual integrated machine, and the TraCI server is started.

4. The traffic optimization evaluation method based on radar-visual integrated machine and AI algorithm according to claim 1, characterized in that, The credit control scheme includes those based on Signal control schemes and adaptive control algorithm schemes for learning algorithms; Wherein, the basis The signal control scheme for the learning algorithm is as follows: Based on state Using the greedy method to select actions And perform the action to receive a reward. The environment has entered a new state. , choose to The largest As To update the value function, the state-action value is updated using the Bellman equation: in, Indicates the learning rate. Indicates the discount factor; The adaptive control algorithm scheme is as follows: During off-peak hours or other periods with fewer vehicles, an adaptive algorithm is used to detect the number of vehicles entering each lane through the integrated radar-visual system and automatically extend the green light time.

5. The traffic optimization evaluation method based on radar-visual integrated machine and AI algorithm according to claim 1, characterized in that, The traffic efficiency indicators include at least average vehicle delay, average travel time, queue length, throughput, average speed, and number of stops.

6. A traffic optimization and evaluation system based on a radar-visual integrated machine and AI algorithms, characterized in that, include: The data acquisition module is used to acquire lidar point cloud and camera image data through the integrated laser vision device, and extract vehicle position, speed, acceleration and trajectory information through a spatiotemporal synchronization mechanism to acquire dynamic traffic flow data on the road in real time, so as to form corresponding multi-source heterogeneous perception data. The data processing module is used to preprocess the multi-source heterogeneous sensing data to generate a high-precision fusion trajectory set; The data simulation module is used to generate road network files, traffic flow files, and simulation configuration files, and to construct a SUMO simulation environment based on the road network files, traffic flow files, and simulation configuration files. The environment simulation module is used to import the high-precision fused trajectory set into the SUMO simulation environment. At the start of the simulation, it generates vehicles according to the vehicle departure time, route, and vehicle type parameters and injects them into the SUMO simulation environment. It outputs the current traffic status data and receives traffic control commands through the TraCI server interface, and continuously records and outputs the simulation status and result data. The strategy generation module is used to read and construct a state vector from the radar-visual integrated machine or the SUMO simulation environment at each control decision, select a traffic control scheme according to the traffic demand of different scenarios and time periods, and generate a traffic control strategy. The comparative analysis module is used to construct a set of schemes to be evaluated. For each scheme, SUMO simulation is run under the same road network, the same traffic demand, and the same evaluation interval. The simulation output results are collected, and traffic efficiency indicators, traffic safety and emission evaluation indicators are calculated. The indicator results of each scheme are normalized and statistically summarized to obtain comparative analysis results between different schemes.

7. The traffic optimization and evaluation system based on radar-visual integrated machine and AI algorithm according to claim 6, characterized in that, The data processing module is specifically used for: The target sequence in the multi-source heterogeneous sensing data is filtered by a combination of rules and statistics, and the trajectory breaks caused by short-term frame loss or occlusion in the multi-source heterogeneous sensing data are completed by a piecewise linear interpolation method. Given a set of sample points sorted in ascending order of the independent variable: interval Then the piecewise linear interpolation is: In the formula, This represents the value of the independent variable to be interpolated. This indicates that when the independent variable is Interpolated estimates at time; The video detection results in the multi-source heterogeneous sensing data are transformed from pixel coordinates to world coordinates, and the radar targets are transformed from polar coordinates to world coordinates, and unified into state variables under the same coordinate system. At the same time, radar tracks and video tracks are correlated to obtain a set of matching pairs. For the matching targets, extended Kalman filtering is used for fusion and updating to obtain a high-precision fused trajectory. For unmatched targets, single-source trajectories are retained and the source and confidence level are marked. For ID switching or broken trajectories, trajectory splicing and reconstruction are performed based on spatiotemporal continuity, lane constraints, and speed trend consistency to form the final trajectory set.

8. The traffic optimization and evaluation system based on radar-visual integrated machine and AI algorithm according to claim 6, characterized in that, The data simulation module is specifically used for: Maps are generated by downloading maps from open-source atlases and then converted into road network files using the netconvert program included in the SUMO simulation. A flow rate file is generated using the multi-source heterogeneous sensing data, and a corresponding simulation configuration file is also generated. A SUMO simulation environment is constructed based on the road network file, the traffic flow file, and the simulation configuration file. Detectors are set in the simulation environment according to the actual deployment location of the radar-visual integrated machine, and the TraCI server is started.

9. A readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the traffic optimization evaluation method based on the radar-visual integrated machine and AI algorithm as described in any one of claims 1 to 5.

10. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the traffic optimization evaluation method based on the radar-visual integrated machine and AI algorithm as described in any one of claims 1 to 5.

Images