A vehicle-road cooperative anti-collision system for underground mine tunnel environment
Through the multi-source data fusion of intelligent road testing units and vehicle-mounted units and V2X communication technology, the blind spots and communication problems of unmanned vehicles in the well industrial and mining tunnel environment are solved, and safe and anti-collision and efficient transportation of underground vehicles are achieved.
Patent Information
- Application Number
- CN202410349128.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-26
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-03-26
AI Technical Summary
The unmanned vehicles of Jing Industrial and Mining have problems such as blind spots, dark light and lots of dust in harsh tunnel environments, resulting in high probability of misidentification, difficult communication signals to penetrate, and difficult to ensure the safe operation of vehicles in the existing technology.
Using intelligent road testing unit and intelligent vehicle-mounted unit, combining V2X communication technology, UWB location service base, mining lidar, camera, millimeter wave radar and V2X roadside equipment, the vehicle position and Real-time monitoring and early warning of obstacle information to make up for the blind spots of perception and ensure smooth communication.
It realizes anti-collision between underground vehicles and between vehicles and obstacles, reduces collision accidents, and improves the safety and efficiency of underground transportation.
Smart Images

Figure CN118254779B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of unmanned driving technology, and in particular to a vehicle-road cooperative collision avoidance system for use in an underground mine tunnel environment. Background Art
[0002] The importance of unmanned driving technology in underground mines cannot be ignored. With the advancement of science and technology, unmanned driving technology has gradually become an important means of safe production in underground mines, bringing many benefits to mines. The unmanned driving collision avoidance system is an important part of unmanned vehicles. It uses sensors and algorithms to achieve safe distance and collision warning between vehicles, effectively reducing the possibility of collision between unmanned vehicles. Unmanned driving collision avoidance systems usually use sensors such as lidar, millimeter wave radar, and cameras to obtain information about the surrounding environment, and use computer vision and image processing technology to process and analyze the collected data to identify surrounding vehicles, pedestrians and other obstacles. At the same time, the system will perform dynamic analysis and prediction based on the vehicle's position, speed and driving trajectory to determine whether there is a potential collision risk. Once a potential collision risk is detected, the unmanned driving collision avoidance system will take appropriate measures to avoid the collision or mitigate the impact of the collision.
[0003] However, unmanned driving in underground mines is still in its early stages, mainly in the following aspects:
[0004] (1) The underground mine environment is harsh, with a large number of irregular tunnels, curved ramps, obstacles, etc., and unmanned vehicles have large blind spots.
[0005] (2) The dark light, high dust and water mist in the mine cause great interference to the perception system of the unmanned vehicle, increasing the probability of misidentification of the unmanned vehicle.
[0006] (3) The underground environment is closed and communication signals are difficult to penetrate. It is necessary to establish a reliable communication system to ensure smooth communication between unmanned vehicles, between vehicles and infrastructure, and between vehicles and ground control centers.
[0007] (4) Mine work involves major safety issues, and the safety requirements for unmanned vehicles are extremely high. Multiple safety measures need to be taken to ensure the safe operation of unmanned vehicles. Therefore, the existing unmanned driving technology is difficult to reuse in underground mines. Summary of the invention
[0008] In view of the above-mentioned deficiencies in the prior art, the present invention provides a vehicle-road collaborative collision avoidance system for use in an underground mine tunnel environment, which can solve the collision avoidance and information interaction problems of unmanned vehicles in the underground mine tunnel environment, and ensure the life safety of miners and the stability of the mine.
[0009] The present invention provides the following technical solution: a vehicle-road cooperative anti-collision system for underground mining tunnel environment, comprising an intelligent road test unit and an intelligent vehicle-mounted unit located underground, and the intelligent road test unit and the intelligent vehicle-mounted unit are connected with a service and management center on the ground through a positioning and communication unit;
[0010] Both the intelligent road test unit and the intelligent vehicle unit use V2X communication technology, which enables the driverless vehicle to obtain information about the surrounding environment and other vehicles in a timely manner while driving;
[0011] The intelligent road test unit includes UWB location service base, mining intelligent microcomputer, mining laser radar, mining camera, mining millimeter wave radar, mining traffic lights and V2X roadside equipment;
[0012] The intelligent road test unit uses more than four UWB location service bases to feed back the positioning results to vehicles or equipment with positioning requirements. The intelligent road test unit uses a mining intelligent microcomputer to accurately control mining laser radar, mining camera, and mining millimeter wave radar to collect tunnel environment data through the drive and corresponding communication interface. It uses the algorithm deployed on the mining intelligent microcomputer to perform fusion detection on multi-source data, and then uses V2X communication technology to send the processed data in real time to nearby vehicles and equipment with real-time communication requirements. The monitoring data and detection results are transmitted to the service and management center on the well through the positioning and communication unit.
[0013] The intelligent vehicle-mounted unit includes UWB positioning card, mining laser radar, mining camera, mining millimeter wave radar, vehicle-mounted domain control host, vehicle-mounted mining power supply, vehicle-mounted switch and V2X vehicle-mounted equipment;
[0014] The intelligent vehicle-mounted unit uses mining laser radar, mining camera, and mining millimeter-wave radar as vehicle-side perception sensors to obtain vehicle coordinate information. The vehicle-mounted domain control host then obtains raw data from the sensor through the vehicle-mounted switch and pre-processes the data.
[0015] The UWB signal sent by the smart vehicle unit through the UWB positioning card is received by the UWB location service base, and the data is transmitted to the computing center for processing through the network layer. Finally, the positioning result is fed back to the smart vehicle unit with positioning requirements, which receives the UWB positioning information through WiFi and integrates it with its own SLAM positioning.
[0016] The intelligent vehicle-mounted unit receives the roadside obstacle information exported by the V2X roadside device in the intelligent road test unit through the V2X vehicle-mounted device, and integrates it with its own perception results to make up for the vehicle's perception blind spots. The vehicle makes control decisions based on the vehicle-road fusion results.
[0017] Preferably, the positioning and communication unit uses optical fiber to open up network channels above and below the well, deploys 5G and 4G base stations underground, and finally sets up wired or wireless WiFi equipment.
[0018] Preferably, multi-source data is fused and detected, such as the following steps:
[0019] S1. Time hard synchronization: Use ptp / gptp to unify the clock source. All sensors use one clock source, with the domain control host as the master and other sensors as slaves, so that subsequent data can be aligned in time dimension.
[0020] S2. Individually sensed target output: Lidar detects target obstacles through top removal, ground removal, and clustering processes. Radar mainly uses ars548 radar to directly output the position and speed information of target obstacles. Camera mainly detects targets through yolov5 deep learning network, mainly including the position and category of the target.
[0021] S3, cache perception results and time soft synchronization: Each node caches the detection results of the last 10 frames, performs time soft synchronization based on the lidar time, and matches the detection results of the radar and camera at the same time;
[0022] S4, spatial synchronization: mainly divided into two steps, first lidar and radar calibration, then lidar and camera calibration, determine the position relationship of lidar, radar and camera and calibrate them into the same spatial coordinate system;
[0023] S5, matching factor calculation: including lidar and radar matching factor calculation and lidar and camera matching factor calculation;
[0024] S6, multi-source data matching: lidar and radar, lidar and camera use Hungarian algorithm for optimal matching;
[0025] S7, target attribute fusion: combining the advantages of lidar, radar and camera, removing redundant information and merging attributes according to different matching situations;
[0026] S8, target tracking: mainly adopts extended Kalman filter to implement the tracking algorithm;
[0027] S9. Fusion result output: The fusion result is converted into a unified message format for publication. The lidar and radar factor calculation process is as follows: calculate the distance L1 from the center point of the lidar obstacle to the center point of the radar obstacle, then calculate the distance L2 from the radar obstacle to the projection point in the direction of the lidar obstacle, and consider L1 and L2 comprehensively to calculate the matching factor: match_score = a*L1+(1-a)*L2, where a is the weight.
[0028] Preferably, the lidar and camera matching factor calculation process is as follows: the obstacle detected by the lidar is projected into the image to form a rectangular frame, and the obstacle detected by the image also forms a rectangular frame, and the ratio between the intersection area and the union area of the two is calculated as the matching factor, that is, the intersection-over-union ratio (IOU).
[0029] Preferably, the target attribute fusion process presents the following relationship: the priority of position fusion is: lidar>radar>camera, and the position information of the matching detection result is based on lidar; the priority of speed fusion is: radar>lidar>camera, and the speed information of the matching detection result is based on radar. The priority of category fusion is: camera>radar>lidar, and the category information of the matching detection result is based on camera.
[0030] Preferably, the extended Kalman filter is to linearize the nonlinear system, use the Taylor expansion, omit the second-order and above terms, obtain an approximate linearized model, and then apply the Kalman filter to complete the estimation. The Taylor expansion formula is:
[0031]
[0032] R n (x) = o[(x-x0) n ].
[0033] Preferably, in the extended Kalman filter, the Taylor expansion retains only the first two terms, that is: f(x)≈f(x0)+f'(x0)*(x-x0). The EKF mainly consists of five equations, which are divided into a prediction equation and an update equation. The prediction equation includes:
[0034] Extrapolate state variables forward:
[0035] Extrapolate the error covariance forward:
[0036] The update equation contains:
[0037] Calculate the extended Kalman filter gain:
[0038] Based on the observed variables, update the estimate:
[0039] Update the error covariance:
[0040] Compared with the prior art, the present invention has the following beneficial effects:
[0041] This vehicle-road collaborative collision avoidance system used in underground mining tunnel environments realizes collision avoidance between underground vehicles, between vehicles and people, and between vehicles and other obstacles. Its main technical features include: real-time monitoring of vehicle position and speed, calculation of obstacle position and speed, calculation of collision risk and issuance of early warning, control of vehicle deceleration or parking, optimization of vehicle collaborative work, etc.; through the application of this device and system, the occurrence of underground vehicle collision accidents can be greatly reduced, and the safety of underground transportation can be improved; at the same time, the system can also optimize the collaborative work of unmanned vehicles and improve the efficiency of underground transportation. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 This is a diagram of the vehicle-road cooperative collision avoidance system of the present invention;
[0043] Figure 2 This is a diagram of the multi-source sensor detection and fusion process of the present invention;
[0044] Figure 3 This is a calculation diagram of lidar and radar matching factors of the present invention;
[0045] Figure 4 This is the lidar and camera matching factor calculation diagram of the present invention. DETAILED DESCRIPTION
[0046] In order to make the objectives, technical solutions and advantages of the embodiments of the present disclosure clearer, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings of the embodiments of the present disclosure. In order to keep the following description of the embodiments of the present disclosure clear and concise, the present disclosure omits detailed descriptions of known functions and known components to avoid unnecessary confusion of the concepts of the present invention.
[0047] See also Figure 1 and Figure 2 A vehicle-road cooperative collision avoidance system for underground mining tunnel environments includes an intelligent road test unit and an intelligent vehicle-mounted unit located underground, and the intelligent road test unit and the intelligent vehicle-mounted unit are connected to a service and management center on the ground through a positioning and communication unit.
[0048] The positioning and communication unit uses optical fiber to open up network channels above and below the well, deploys 5G and 4G base stations underground, and finally sets up wired or wireless WiFi equipment.
[0049] First, use optical fiber to open up the network channel above and below the well, and use the high bandwidth and low loss characteristics of optical fiber to ensure stable and reliable data transmission above and below the well. Then, deploy 5G and 4G base stations underground, use the high-speed data transmission and large capacity characteristics of 5G to meet the data communication needs of high-density equipment underground, and use the 4G network to provide stable coverage to ensure the continuity of communication underground. Finally, install wired or wireless WiFi equipment to expand the coverage of the underground network, so that underground workers and equipment can access the network and realize real-time data transmission and remote control.
[0050] V2X communication technology is used in both the intelligent road test unit and the intelligent vehicle unit. V2X technology can provide highly reliable and low-latency communication, enabling unmanned vehicles to obtain information about the surrounding environment and other vehicles in a timely manner while driving.
[0051] The intelligent road test unit includes UWB location service base, mining intelligent microcomputer, mining laser radar, mining camera, mining millimeter wave radar, mining traffic lights and V2X roadside equipment;
[0052] The intelligent road test unit uses more than four UWB location service bases to feed back the positioning results to vehicles or equipment with positioning requirements. The intelligent road test unit uses a mining intelligent microcomputer to accurately control mining laser radars, mining cameras, and mining millimeter-wave radars to collect tunnel environment data through drivers and corresponding communication interfaces, especially in complex areas such as mine intersections, uphill and downhill slopes, and chambers. These places often have perception blind spots. The algorithm deployed on the microcomputer is used to fuse and detect multi-source data, effectively reducing the impact of adverse factors such as dark light, dust, and water mist.
[0053] The multi-source data is fused and detected through the algorithm deployed on the mining intelligent microcomputer, and then the processed data is sent in real time to nearby vehicles and equipment with real-time communication needs using V2X communication technology. The monitoring data and detection results are transmitted to the service and management center on the well through the positioning and communication unit.
[0054] Multi-source data is fused for detection, as shown in the following steps: where lidar represents mining laser radar, radar represents mining millimeter-wave radar, and camera represents mining camera.
[0055] S1. Time hard synchronization: Use ptp / gptp to unify the clock source. All sensors use one clock source, with the domain control host as the master and other sensors as slaves, so that subsequent data can be aligned in time dimension.
[0056] S2. Individually sensed target output: Lidar detects target obstacles through top removal, ground removal, and clustering processes. Radar mainly uses ars548 radar to directly output the position and speed information of target obstacles. Camera mainly detects targets through yolov5 deep learning network, mainly including the position and category of the target.
[0057] S3, cache perception results and time soft synchronization: Each node caches the detection results of the last 10 frames, performs time soft synchronization based on the lidar time, and matches the detection results of the radar and camera at the same time;
[0058] S4, spatial synchronization: mainly divided into two steps, first lidar and radar calibration, then lidar and camera calibration, determine the position relationship of lidar, radar and camera and calibrate them into the same spatial coordinate system;
[0059] S5, matching factor calculation: including lidar and radar matching factor calculation and lidar and camera matching factor calculation;
[0060] Figure 3 It can be seen that the calculation process of lidar and radar factors is: calculate the distance L1 from the center point of the lidar obstacle to the center point of the radar obstacle, then calculate the distance L2 from the radar obstacle to the projection point in the direction of the lidar obstacle, and comprehensively consider L1 and L2 to calculate the matching factor: match_score = a*L1+(1-a)*L2, a is the weight. The matching factor comprehensively considers the two factors of distance and angle. The closer the distance, the smaller L1, and the smaller the angle, the smaller L2.
[0061] Figure 4 It can be seen that the calculation process of the lidar and camera matching factor is as follows: the obstacle detected by the lidar is projected into the image to form a rectangular frame, and the obstacle detected by the image also has a rectangular frame. The ratio between the intersection area and the union area of the two is calculated as the matching factor, which is the intersection-over-union (IOU) ratio.
[0062] S6. Multi-source data matching: lidar and radar, lidar and camera are matched optimally using the Hungarian algorithm.
[0063] S7, target attribute fusion: Combining the advantages of lidar, radar and camera, according to different matching situations, remove redundant information and merge attributes.
[0064] The target attribute fusion process presents the following relationship: the priority of position fusion: lidar>radar>camera, the position information of the matching detection result is based on lidar; the priority of speed fusion: radar>lidar>camera, the speed information of the matching detection result is based on radar. The priority of category fusion: camera>radar>lidar, the category information of the matching detection result is based on camera.
[0065] S8, target tracking: mainly adopts extended Kalman filter to implement the tracking algorithm;
[0066] Kalman filtering can provide a good estimate of the target state in a linear model with a Gaussian error model, but if there are nonlinear factors in the system, the effect is not ideal. Typical nonlinear function relationships include square relationships, logarithmic relationships, exponential relationships, trigonometric function relationships, etc. For nonlinear system problems, the more commonly used method is to linearize the nonlinear part. The extended Kalman filter is to linearize the nonlinear system, use the Taylor expansion, ignore the second-order and above terms, and obtain an approximate linearized model. Then, the Kalman filter is applied to complete the estimation. The Taylor expansion formula is:
[0067]
[0068] R n (x) = o[(x-x0) n ].
[0069] In the extended Kalman filter, the Taylor expansion only retains the first two terms, that is: f(x)≈f(x0)+f'(x0)*(x-x0). The EKF mainly consists of five equations, which are divided into prediction equations and update equations. The prediction equation contains:
[0070] Extrapolate state variables forward:
[0071] Extrapolate the error covariance forward:
[0072] The update equation contains:
[0073] Calculate the extended Kalman filter gain:
[0074] Based on the observed variables, update the estimate:
[0075] Update the error covariance:
[0076] S9. Fusion result output: The fusion results are converted into a unified message format for publication.
[0077] The intelligent vehicle-mounted unit includes UWB positioning card, mining laser radar, mining camera, mining millimeter wave radar, vehicle-mounted domain control host, vehicle-mounted mining power supply, vehicle-mounted switch and V2X vehicle-mounted equipment;
[0078] The intelligent vehicle-mounted unit uses mining laser radar, mining camera, and mining millimeter-wave radar as vehicle-side perception sensors to obtain vehicle coordinate information;
[0079] When deploying sensors, in order to ensure that there are no blind spots, especially in the low-speed environment underground, in order to achieve this goal, we need to ensure that the front distance coverage is at least 80 meters, the rear coverage is at least 50 meters, and the coverage on both sides is at least 30 meters. However, a single radar may be difficult to cover all areas, so we need to use a top long-range lidar and four blind-filling lidars in the front, back, left and right to achieve more comprehensive coverage. At the same time, we also need to use millimeter-wave radars and cameras to further enhance perception capabilities. A front millimeter-wave radar and three blind-filling millimeter-wave radars are used for range coverage, while a front camera focuses on detecting the category information of obstacles in front. In this way, we can obtain more accurate environmental information and provide strong support for subsequent decision-making and control.
[0080] Sensor calibration is a key step to ensure data accuracy. We first accurately calibrate the lidar and vehicle position to ensure that the sensor data is aligned with the vehicle coordinate system. Then, we jointly calibrate different lidars to ensure that the data between them can accurately match. In addition, we also jointly calibrate lidar with millimeter wave radar and lidar with camera to achieve data alignment between multiple sensors. Through these calibration steps, we can obtain the position information of all sensors relative to the vehicle coordinate system, providing a reliable benchmark for subsequent data processing and fusion.
[0081] The vehicle-mounted domain control host then obtains raw data from the sensor through the vehicle-mounted switch and preprocesses the data, including denoising, filtering, spatial alignment, and temporal alignment operations, to improve the accuracy and reliability of the data. The preprocessed data provides high-quality input for subsequent algorithm detection and fusion.
[0082] The UWB signal sent by the smart vehicle unit through the UWB positioning card is received by the UWB location service base, and the data is transmitted to the computing center for processing through the network layer. Finally, the positioning result is fed back to the smart vehicle unit with positioning requirements, which receives the UWB positioning information through WiFi and integrates it with its own SLAM positioning.
[0083] The fusion process is as follows:
[0084] UWB positioning information acquisition: Receive feedback data from the computing center in real time via WiFi to obtain the vehicle's UWB positioning information.
[0085] UWB positioning data preprocessing: The collected UWB signals include signal denoising and data filtering to improve the accuracy and reliability of the data.
[0086] SLAM positioning: Using the odometer, wheel speed sensor, IMU and lidar data, the relative position and posture information of the vehicle is calculated through the SLAM algorithm.
[0087] Data fusion: Using UWB and SLAM positioning information, the Extended Kalman Filter (EKF) algorithm is used to fuse the two data to improve the accuracy and stability of vehicle positioning. The main change here is to use the UWB and SLAM positioning results at different times as observation values to update the equation. Finally, the positioning results of the two are combined to estimate relatively accurate position information.
[0088] The intelligent vehicle-mounted unit receives the roadside obstacle information exported by the V2X roadside device in the intelligent road test unit through the V2X vehicle-mounted device, and integrates it with its own perception results to make up for the vehicle's perception blind spots. The vehicle makes control decisions based on the vehicle-road fusion results.
[0089] The perception results are fused as follows:
[0090] Data reception: Through V2X technology, the vehicle-mounted unit can receive obstacle information sent from roadside equipment, including the location, speed, and size of the obstacle.
[0091] Data preprocessing: Preprocess the received roadside obstacle data, including data decoding, denoising, and filtering operations to improve the accuracy and reliability of the data.
[0092] Self-perception: Use on-board sensors to perceive the environment around the vehicle and extract information about the location, speed, and size of obstacles. This information can be recorded and processed for integration with roadside obstacle information.
[0093] Data fusion: Fusion of obstacle information from the roadside and the obstacle information perceived by the vehicle. Since both the vehicle-side and the roadside are fused perception results, the probability that the obstacles output by both are real obstacles is relatively high. The fusion here adopts the union method. First, Hungarian matching is performed on obstacles with similar positions; then, obstacles without matching are retained as blind obstacles for each other; finally, the matched obstacles are based on the obstacles detected in real time by the vehicle-side.
[0094] The process of the vehicle making control decisions based on the vehicle-road fusion results is as follows:
[0095] Decision-making and planning: Decision-making and planning are carried out based on the integrated environmental information. The decision-making system will formulate corresponding control strategies according to the current road conditions, traffic conditions, and vehicle status information. This step usually involves the application of algorithms such as path planning and behavior prediction.
[0096] Control execution: Based on the output of the decision system, the vehicle's control system will perform corresponding actions according to the decision results. This includes steering, acceleration, deceleration, braking, and control of in-vehicle equipment such as lights, wipers, and air conditioning. The execution layer is responsible for converting these instructions into specific driving actions and accurately controlling various parts of the vehicle through the bus system.
[0097] Feedback adjustment: During the driving process of the vehicle, the decision-making system will continuously receive new environmental information and adjust the control strategy in a timely manner through the feedback loop to ensure that the vehicle is always in a safe and stable state.
[0098] The above embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention. The protection scope of the present invention is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the essence and protection scope of the present invention, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present invention.
Claims
1. A vehicle-road cooperative anti-collision system for underground mining tunnel environments. Features: It includes an intelligent road test unit and an intelligent vehicle unit located underground, and the intelligent road test unit and the intelligent vehicle unit are connected with the service and management center on the ground through the positioning and communication unit; Both the intelligent road test unit and the intelligent vehicle unit use V2X communication technology, which enables the driverless vehicle to obtain information about the surrounding environment and other vehicles in a timely manner while driving; The intelligent road test unit includes UWB location service base, mining intelligent microcomputer, mining laser radar, mining camera, mining millimeter wave radar, mining traffic lights and V2X roadside equipment; The intelligent road test unit uses more than four UWB location service bases to feed back the positioning results to vehicles or equipment with positioning requirements. The intelligent road test unit uses a mining intelligent microcomputer to accurately control mining laser radar, mining camera, and mining millimeter wave radar to collect tunnel environment data through the drive and corresponding communication interface. It uses the algorithm deployed on the mining intelligent microcomputer to perform fusion detection on multi-source data, and then uses V2X communication technology to send the processed data in real time to nearby vehicles and equipment with real-time communication requirements. The monitoring data and detection results are transmitted to the service and management center on the well through the positioning and communication unit. The intelligent vehicle-mounted unit includes UWB positioning card, mining laser radar, mining camera, mining millimeter wave radar, vehicle-mounted domain control host, vehicle-mounted mining power supply, vehicle-mounted switch and V2X vehicle-mounted equipment; The intelligent vehicle-mounted unit uses mining laser radar, mining camera, and mining millimeter-wave radar as vehicle-side perception sensors to obtain vehicle coordinate information. The vehicle-mounted domain control host then obtains raw data from the sensor through the vehicle-mounted switch and pre-processes the data. The UWB signal sent by the smart vehicle unit through the UWB positioning card is received by the UWB location service base, and the data is transmitted to the computing center for processing through the network layer. Finally, the positioning result is fed back to the smart vehicle unit with positioning requirements, which receives the UWB positioning information through WiFi and integrates it with its own SLAM positioning. The intelligent vehicle unit receives the roadside obstacle information derived from the V2X roadside device in the intelligent road test unit through the V2X vehicle-mounted device, and integrates it with its own perception results to make up for the vehicle's perception blind spot. The vehicle makes control decisions based on the vehicle-road fusion results. The multi-source data is fused for detection, as shown in the following steps, where lidar represents mining laser radar, radar represents mining millimeter-wave radar, and camera represents mining camera: S1. Time hard synchronization: Use ptp / gptp to unify the clock source. All sensors use one clock source, with the domain control host as the master and other sensors as slaves, so that subsequent data can be aligned in time dimension. S2. Individually sensed target output: Lidar detects target obstacles through top removal, ground removal, and clustering processes. Radar mainly uses ars548 radar to directly output the position and speed information of target obstacles. Camera mainly detects targets through yolov5 deep learning network, mainly including the position and category of the target. S3, cache perception results and time soft synchronization: Each node caches the detection results of the last 10 frames, performs time soft synchronization based on the lidar time, and matches the detection results of the radar and camera at the same time; S4, spatial synchronization: mainly divided into two steps, first lidar and radar calibration, then lidar and camera calibration, determine the position relationship of lidar, radar and camera and calibrate them into the same spatial coordinate system; S5, matching factor calculation: including lidar and radar matching factor calculation and lidar and camera matching factor calculation; S6, multi-source data matching: lidar and radar, lidar and camera use Hungarian algorithm for optimal matching; S7, target attribute fusion: combining the advantages of lidar, radar and camera, removing redundant information and merging attributes according to different matching situations; S8, target tracking: mainly adopts extended Kalman filter to implement the tracking algorithm; S9. Fusion result output: The fusion results are converted into a unified message format for publication.
2. A vehicle-road cooperative collision avoidance system for use in an underground mine tunnel environment according to claim 1, Features: The positioning and communication unit uses optical fiber to open up network channels above and below the well, deploys 5G and 4G base stations underground, and finally sets up wired or wireless WiFi equipment.
3. A vehicle-road cooperative collision avoidance system for use in an underground mine tunnel environment according to claim 1, It is characterized in that The calculation process of lidar and radar factors is as follows: calculate the distance L1 from the center point of the lidar obstacle to the center point of the radar obstacle, then calculate the distance L2 from the radar obstacle to the projection point in the direction of the lidar obstacle, and comprehensively consider L1 and L2 to calculate the matching factor: match_score = a*L1 + (1-a)*L2, where a is the weight.
4. A vehicle-road cooperative collision avoidance system for use in an underground mine tunnel environment according to claim 1, It is characterized in that The calculation process of lidar and camera matching factor is as follows: the obstacle detected by lidar is projected into the image to form a rectangular frame, and the obstacle detected by the image also forms a rectangular frame. The ratio between the intersection area and the union area of the two is calculated as the matching factor, which is the intersection-union ratio.
5. The vehicle-road cooperative anti-collision system for underground mining tunnel environment according to claim 1, Features: The target attribute fusion process presents the following relationship: the priority of position fusion is: lidar>radar>camera, and the position information of the matching detection result is based on lidar; the priority of speed fusion is: radar>lidar>camera, and the speed information of the matching detection result is based on radar; the priority of category fusion is: camera>radar>lidar, and the category information of the matching detection result is based on camera.
6. A vehicle-road cooperative collision avoidance system for use in an underground mine tunnel environment according to claim 1, Features: The extended Kalman filter is to linearize the nonlinear system, use the Taylor expansion, ignore the second-order and above terms, get an approximate linearized model, and then apply the Kalman filter to complete the estimation. The Taylor expansion formula is: ; 。 7. A vehicle-road cooperative collision avoidance system for use in an underground mine tunnel environment according to claim 1, Features: In the extended Kalman filter, the Taylor expansion retains only the first two terms, namely: , EKF mainly consists of five equations, which are divided into prediction equation and update equation. The prediction equation includes: Extrapolate state variables forward: ; Extrapolate the error covariance forward: ; The update equation contains: Calculate the extended Kalman filter gain: ; Based on the observed variables, update the estimate: ; Update the error covariance: .
Citation Information
Patent Citations
Digital twin transportation system and method for underground unmanned trackless vehicle
CN117307240A
Unmanned vehicle control method based on V2I technology in well mining scene
CN117406747A
