A road safety device and method for a V2X-based unmanned driving system
By deploying roadside sensing devices and V2X communication technology at mine roadway intersections, global data at the intersections is acquired and combined with roadway geometric boundary constraints, solving the problem of blind spots in the perception of unmanned vehicles in the mine environment. This enables beyond-line-of-sight perception and safety control, significantly improving the safety of unmanned driving in the mine environment.
Patent Information
- Application Number
- CN202511234593.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-01
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2045-09-01
AI Technical Summary
In mining environments, unmanned vehicles face blind spots due to obstructions from tunnel walls, making it difficult to detect vehicles from other directions at intersections and resulting in a high risk of collisions. Existing V2X technology in mining scenarios lacks blind spot identification and supplementation mechanisms and does not consider the impact of tunnel geometry constraints on vehicle trajectory prediction, making it difficult to fully realize its safety guarantee role.
By deploying roadside sensing devices and computing units at alleyway intersections, V2X communication technology is used to acquire global data at the intersections and broadcast it to autonomous vehicles. Combined with alleyway geometric boundary constraints and increasing the safe distance threshold for blind spot targets, the vehicle's beyond-line-of-sight perception and safety control are achieved.
It effectively solves the problem of blind spots for vehicle sensors caused by tunnel walls in mining environments, significantly improves the safety and collision avoidance decision-making capabilities of autonomous vehicles, and reduces the risk of collisions.
Smart Images

Figure CN121122047B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of vehicle positioning, and in particular to a road safety device and method for an unmanned driving system based on V2X. Background Technology
[0002] With the rapid development of intelligent mining technology, the application of unmanned driving technology in enclosed environments such as underground mines and tunnels is becoming increasingly widespread. Compared with open roads on the ground, the mine environment is characterized by narrow spaces, poor lighting conditions, and winding tunnels, which places higher demands on the environmental perception and safety control of unmanned vehicles.
[0003] Mine tunnels present numerous unique environments, including chambers, tunnel walls, and narrow intersections. Particularly at tunnel intersections, the obstruction of tunnel walls and the limitation of road width severely restrict the sensing range of vehicle-mounted sensors (such as lidar, millimeter-wave radar, and cameras), making it impossible to effectively detect vehicles from other tunnels. This blind spot exposes autonomous vehicles to an extremely high risk of collision when crossing intersections, seriously impacting the safety and efficiency of mine transportation.
[0004] Existing vehicle-mounted perception systems primarily rely on single-vehicle intelligence, meaning they perceive the environment solely through the sensors mounted on the vehicle itself. However, in the environment of mine tunnels, even with high-performance LiDAR and millimeter-wave radar, the detection range of onboard sensors still cannot penetrate the tunnel walls, resulting in almost zero perception capability for vehicles approaching from other directions at intersections. This "perception island" phenomenon means that autonomous vehicles can only detect vehicles approaching from the side after entering the intersection, leaving an extremely limited time window for collision avoidance decisions and execution.
[0005] While V2X (Vehicle-to-Everything) communication technology offers a solution to beyond-line-of-sight perception problems by deploying roadside sensing and communication equipment at intersections to acquire and transmit global traffic information, existing V2X technologies still have several shortcomings in mine environments: a lack of blind spot identification and supplementation mechanisms for tunnel environments; failure to consider the impact of tunnel geometric constraints on vehicle trajectory prediction; and a lack of differentiated risk assessment strategies for targets in blind spots. These issues prevent existing solutions from fully leveraging the safety assurance role of V2X technology in mine environments. Summary of the Invention
[0006] To address the blind spot problem of vehicle-mounted sensors caused by tunnel walls obstructing vision in mining environments, this application provides a road safety device and method for an unmanned driving system based on V2X. By acquiring intersection data through roadside sensing equipment groups as supplementary data for blind spots, the communication security of unmanned vehicles in mining environments is improved.
[0007] One aspect of this application provides a road safety device for a V2X-based unmanned driving system, comprising: a roadside system installed at a road intersection, including: a roadside computing unit (MEC) for fusing and calculating data collected by roadside sensing devices; the roadside computing unit (MEC) broadcasting the fusion calculation results via a 5.8GHz frequency band signal and receiving signals of the same frequency band from vehicle-mounted communication devices; and a roadside sensing device group including a lidar, millimeter-wave radar, and a camera installed on the roadside for real-time sensing of obstacles and passing vehicle data in the intersection area.
[0008] The vehicle-mounted system, installed on the autonomous vehicle, includes: an onboard unit (OBU) that broadcasts vehicle data and receives broadcast data from a roadside unit (RSU); an autonomous vehicle-mounted controller that receives environmental data from the vehicle-mounted sensing device group and roadside data received by the OBU, performs fusion calculations, and then controls the vehicle's steering and speed; and a vehicle-mounted sensing device group that includes lidar, millimeter-wave radar, and cameras installed on the autonomous vehicle.
[0009] The roadside system acquires global data of the intersection area through perception fusion and transmits the global data to the autonomous vehicles entering the intersection area through V2X communication, so that the autonomous vehicles can obtain beyond-line-of-sight perception capabilities.
[0010] In this solution, V2X (Vehicle-to-Everything) refers to the wireless communication technology between unmanned vehicles and roadside infrastructure in a mining environment. Through V2X communication, roadside sensing devices deployed at the entrance of the tunnel can transmit the detected vehicle and obstacle information in real time to the unmanned vehicle about to enter the tunnel, realizing information sharing between the road and the vehicle, thereby overcoming the physical obstruction of the tunnel walls to the vehicle-mounted sensors.
[0011] Mobile Edge Computing (MEC): In this solution, the MEC is an edge computing device deployed at the intersection of alleyways. It is responsible for real-time processing and fusion computing of raw data collected by roadside LiDAR, millimeter-wave radar, and cameras. The MEC can complete calculation tasks such as target detection, tracking, and state estimation locally on the roadside. It broadcasts the processed structured data (such as vehicle position, speed, and heading) through the RSU, avoiding the transmission of large amounts of raw sensor data and ensuring real-time communication.
[0012] Roadside Unit (RSU): In this scheme, the RSU is a dedicated short-range communication device installed at the intersection of lane roads, serving as a communication bridge between the roadside system and the vehicle-mounted system. The RSU is responsible for broadcasting the intersection global perception data processed by the MEC to all vehicles in the intersection area via the 5.8GHz frequency band, while simultaneously receiving vehicle status information sent by the vehicle's OBU, thus realizing two-way communication between the road and the vehicle.
[0013] 5.8GHz band signal: In this solution, the 5.8GHz band is a dedicated short-range communication (DSRC) band specifically allocated to intelligent transportation systems, characterized by low transmission latency (typically less than 100ms) and high reliability. In the enclosed environment of a mine, this band is less susceptible to external interference, ensuring timely and stable transmission of roadside sensing data to vehicles, meeting the stringent real-time communication requirements for vehicle collision avoidance decisions.
[0014] On-Board Unit (OBU): In this solution, the OBU is a communication device installed on the autonomous vehicle and used in conjunction with the Roadside Unit (RSU). The OBU receives global perception data of the intersection broadcast by the RSU and transmits it to the vehicle-side controller for fusion processing. Simultaneously, it broadcasts the vehicle's position, speed, heading, and other status information, enabling the roadside system and other vehicles to perceive the vehicle's motion status and achieve comprehensive information sharing.
[0015] Beyond-line-of-sight perception capability: In this solution, beyond-line-of-sight perception refers to the ability of autonomous vehicles to obtain environmental perception beyond the physical detection range of their own sensors by receiving roadside perception data. Especially at mine tunnel intersections, onboard sensors are unable to detect vehicles approaching from the opposite direction due to tunnel walls. However, roadside perception data acquired through V2X can provide target information within these blind spots, enabling vehicles to perceive potential risks in obscured areas in advance, thus enhancing the ability to go from "unseen" to "visible."
[0016] Another aspect of this application provides a road safety control method for a V2X-based unmanned driving system, comprising: S1, collecting environmental data of the intersection area through a roadside perception device group, the environmental data including lidar data, millimeter-wave radar data, and image data; S2, using a roadside computing unit (MEC) to fuse the collected lidar data, millimeter-wave radar data, and image data to obtain global intersection data; the global intersection data includes: the position, speed, and heading of all vehicles within the intersection, as well as the position and size of obstacles in the intersection area; S3, broadcasting the global intersection data via a roadside communication unit (RSU) through the 5.8 GHz frequency band; S4, The autonomous vehicle receives global intersection data broadcast by the roadside via its On-Board Unit (OBU); S5, the autonomous vehicle controller fuses the received global intersection data with the vehicle-side environmental data collected by the vehicle-side perception equipment group to obtain fused perception data; the fused perception data includes data within the autonomous vehicle's own perception range and data within the blind spots of its own sensors; S6, the autonomous vehicle controller determines whether there is a collision risk between the vehicle and other vehicles or obstacles in the intersection area based on the fused perception data; S7, when a collision risk is detected, the autonomous vehicle controller controls the autonomous vehicle to accelerate or decelerate to avoid a collision.
[0017] Further, in step S2, global intersection data is obtained, including: filtering the lidar data and extracting point cloud clusters using a clustering algorithm; distinguishing static and dynamic targets based on the point cloud clusters and calculating the three-dimensional coordinates of each target in the roadside coordinate system; performing a fast Fourier transform on the millimeter-wave radar data to calculate the radial velocity of each dynamic target; obtaining the target category and direction of motion using a target detection algorithm based on the image data, where the target category includes vehicles and obstacles; associating the obtained target three-dimensional coordinates, dynamic target radial velocity, target category, and direction of motion to obtain target data; and constructing global intersection data based on the target data.
[0018] Furthermore, based on the target data, global intersection data is constructed, including: dividing the target data into vehicle target sets and obstacle target sets according to target categories; generating vehicle state data for each vehicle target in the vehicle target set: using the vehicle target's three-dimensional coordinates as the vehicle position; calculating the corresponding vehicle target's velocity vector in the roadside coordinate system based on the vehicle target's radial velocity and direction of motion, as the vehicle velocity; and using the vehicle target's direction of motion as the vehicle heading; generating obstacle state data for each obstacle target in the obstacle target set: using the obstacle target's three-dimensional coordinates as the obstacle position; calculating the obstacle's length, width, and height based on the boundary of the point cloud cluster corresponding to the obstacle target, as the obstacle dimensions; and outputting the vehicle state data and obstacle state data as global intersection data.
[0019] Furthermore, S5, obtains fused perception data, including: extracting the position coordinates of vehicle targets based on vehicle-side environmental data collected by the vehicle-side perception device group. Based on speed and category, construct a set of vehicle-side perceived targets; based on the current position coordinates of the autonomous vehicle... and heading angle Based on the parameters of the vehicle-end perception device group, the boundary of the vehicle-end perception area is determined, and the area outside the boundary of the vehicle-end perception area is defined as the vehicle-end perception blind zone. The global data of the intersection is transformed from the roadside coordinate system to the vehicle coordinate system centered on the autonomous vehicle to obtain the roadside perception target set in the vehicle coordinate system. Target matching is performed on the vehicle-end perception target set and the roadside perception target set in the vehicle coordinate system to obtain fused perception data. The fused perception data includes: targets in the vehicle-end perception target set, and roadside perception targets located in the vehicle-end perception blind zone.
[0020] Furthermore, the boundary of the vehicle-end perception area is determined, and the area outside the boundary of the vehicle-end perception area is defined as the vehicle-end perception blind zone. This includes: acquiring the detection parameters of the vehicle-end perception device group, including the maximum detection range and field of view of the vehicle-mounted LiDAR, the maximum detection range and detection angle of the vehicle-mounted millimeter-wave radar, and the effective recognition range and field of view of the vehicle-mounted camera; determining the perception area of each sensor based on the detection parameters of each sensor, and performing a union operation to obtain the comprehensive perception area; acquiring the position of the tunnel wall, calculating the occlusion of the tunnel wall on the comprehensive perception area, and removing the occluded area from the comprehensive perception area; and using the boundary of the comprehensive perception area after removing the occlusion as the boundary of the vehicle-end perception area.
[0021] Furthermore, the fused perception data includes: spatially associating the vehicle-side perceived target set and the roadside perceived target set in the vehicle coordinate system, and calculating the positional distance between the targets; determining candidate matching target pairs based on the positional distance and a preset matching threshold; verifying the motion feature similarity of the candidate matching target pairs, confirming the target pairs that meet the similarity conditions as the same target, and using the vehicle-side perceived data as the fused data for the matching targets; determining whether the unmatched targets in the roadside perceived target set are located within the determined vehicle-side perceived blind zone, and identifying targets located within the blind zone as blind zone supplementary targets; merging the vehicle-side perceived target set and the blind zone supplementary targets to generate fused perception data containing targets within the vehicle-side perceived range and targets within the perceived blind zone.
[0022] Furthermore, in S6, the autonomous vehicle's on-board controller, based on fused perception data, determines whether there is a risk of collision between the vehicle and other vehicles or obstacles in the intersection area. This includes: extracting the position, speed, category, and data source identifier of all targets from the fused perception data, where targets include vehicle targets and obstacle targets; and obtaining the current position of the autonomous vehicle. Current speed and heading angle According to the preset time interval Generating future time series ,in, Based on the vehicle kinematics model, predict the autonomous vehicle at each moment. Location The predicted trajectory of this vehicle is obtained. For each vehicle target j in the fused sensing data, extract its current position. ,speed and heading angle Based on the future time series T, predict the vehicle target j at each time step. Location The predicted trajectory of vehicle target j is obtained. For vehicle target j whose data source is identified as a blind spot in vehicle-side perception, the corresponding predicted trajectory is obtained using the geometric boundary data of the roadway. Apply constraints; based on the predicted trajectory of this vehicle. Calculate whether there is a risk of collision between the autonomous vehicle i and other vehicles or obstacles in the intersection area.
[0023] Furthermore, for vehicle target j whose data source is identified as a blind spot in vehicle-side perception, the corresponding predicted trajectory is obtained using the geometric boundary data of the roadway. Constraints are applied, including: acquiring the geometric boundary data of the roadway to determine the drivable area of the roadway; and determining the predicted trajectory of vehicle target j within the blind spot of vehicle-end perception. The system checks whether each predicted position is within the drivable area of the roadway. For predicted positions that are outside the drivable area, the positions are corrected according to the roadway boundary constraints to ensure that the corrected positions are within the drivable area, thus obtaining the constrained predicted trajectory.
[0024] Furthermore, based on the predicted trajectory of this vehicle... Calculate whether there is a collision risk between the autonomous vehicle i and other vehicles or obstacles in the intersection area, including: for each autonomous vehicle i, at each time step Calculate the predicted position of this vehicle Euclidean distance between the predicted position of vehicle target j and the target j: If the data source of vehicle target j is identified as the vehicle-side perception blind spot, the predicted position after constraints is used. Calculate distance If the data source for vehicle target j is identified as vehicle-side perception, the original predicted location is used. Calculate distance For each obstacle target k, obtain its position and size information, and construct the boundary rectangle of the obstacle; at each time step... Calculate the predicted position of this vehicle Minimum distance to the boundary rectangle of obstacle k Set a safe distance threshold for vehicle targets. Set a safe distance threshold for obstacle targets. For targets originating from blind spots in vehicle-mounted perception, the corresponding safe distance threshold is increased by a factor of α, where α > 1; when there exists any moment... , making or At that time, it was determined that there was a risk of collision;
[0025] Output the collision risk assessment results, including whether there is a collision risk, the minimum distance value, and the corresponding target identifier.
[0026] Compared to existing technologies, the advantages of this application are:
[0027] This application utilizes V2X communication technology to achieve real-time fusion of roadside perception data and vehicle-mounted perception data, effectively solving the problem of blind spots in vehicle-mounted sensor perception caused by tunnel walls obstructing the view in a mining environment. The roadside perception equipment group is deployed at a high point at the intersection, enabling it to acquire vehicle and obstacle information from a global perspective. This information is broadcast in real-time to autonomous vehicles entering the intersection via the 5.8GHz frequency band. After receiving the roadside data, the vehicle-mounted controller accurately identifies and marks targets located within the vehicle-mounted perception blind spot as "blind spot supplementary targets" through coordinate transformation and target matching, thereby enabling the autonomous vehicle to detect oncoming vehicles obstructed by tunnel walls.
[0028] Furthermore, this application, on the one hand, utilizes the geometric boundaries of the roadway to constrain the predicted trajectory of vehicles in the blind spot, avoiding trajectory prediction deviations caused by the lack of direct observation; on the other hand, it increases the safe distance threshold for targets in the blind spot by a factor of α (α>1), reserving a larger safety margin for perception uncertainties. This application fundamentally eliminates the collision risk caused by perception blind spots, significantly improving the safety of unmanned driving in mining environments. Attached Figure Description
[0029] This application will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:
[0030] Figure 1 This is a schematic diagram of a road safety device for a V2X-based unmanned driving system, according to some embodiments of this application.
[0031] Figure 2 This is a layout diagram of the vehicle-side hardware device according to some embodiments of this application;
[0032] Figure 3 This is a layout diagram of a roadside device according to some embodiments of this application;
[0033] Figure 4 This is an exemplary flowchart illustrating a road safety method for a V2X-based unmanned driving system according to some embodiments of this application. Detailed Implementation
[0034] The methods and systems provided in the embodiments of this application will now be described in detail with reference to the accompanying drawings.
[0035] Example 1
[0036] Hardware system architecture and included equipment of V2X-based autonomous driving system road safety devices and systems, such as Figure 1 As shown: Roadside Computing Unit (MEC): An edge computing device installed on the roadside, its main function is to perform fusion calculations on data collected by roadside sensing devices. Roadside Communication Unit (RSU): A communication unit installed on the roadside, its main function is to broadcast the data fused and calculated by the roadside edge computing device via a 5.8GHz frequency band signal; and to receive signals in the same frequency band from vehicle-mounted communication devices.
[0037] Onboard Unit (OBU): The onboard unit is generally installed in autonomous vehicles and other auxiliary equipment. The OBU is responsible for broadcasting vehicle information and receiving broadcast information from the Roadside Unit (RSU).
[0038] Visual sensors – cameras: visual sensors are used to acquire image information of the environment and provide visual image data to the perception module in autonomous driving; LiDAR sensors – LiDAR: LiDAR sensors are used to acquire three-dimensional information of the environment and provide data to the perception and positioning modules in autonomous driving; Millimeter-wave sensors – millimeter-wave radar: millimeter-wave radar sensors are used to acquire speed information of other road users and provide data to the perception module in autonomous driving.
[0039] Autonomous driving vehicle controller: Installed on the vehicle-mounted vehicle, it can receive information about the vehicle's surrounding environment from the vehicle's perception sensors, and fuse and calculate information about other traffic participants obtained from the perception, thereby controlling the vehicle's steering and speed, so that the vehicle can travel safely and reliably along the predetermined trajectory;
[0040] Installation diagram and functions of vehicle-mounted equipment for road safety devices in V2X-based autonomous driving systems, as shown below. Figure 2 As shown:
[0041] The vehicle is equipped with a lidar 1 mounted on the roof; a camera 2 mounted inside the windshield of the driver's cab; an autonomous driving vehicle-side controller 3; a millimeter-wave radar 4 mounted on the front of the vehicle; an onboard unit (OBU) 5 mounted on the vehicle; and a vehicle 6 with a drive-by-wire chassis. Through the perception fusion of lidar 1, camera 2, and millimeter-wave radar 4, the vehicle can perceive obstacles and other vehicles on the planned path in real time and broadcast the information through the onboard unit (OBU) 5 to synchronize the information with the roadside unit (RSU) and the OBU devices of other vehicles.
[0042] Installation diagram and functions of roadside end equipment for V2X-based autonomous driving systems, as shown in the diagram. Figure 3 As shown:
[0043] The roadside computing unit MEC7 is installed on the roadside; the roadside communication unit RSU8 is installed at the roadside end; the lidar 9 is installed on the roadside; the millimeter-wave radar 10 is installed on the roadside; and the camera 11 is installed on the roadside.
[0044] In practical work, through the perception fusion of 9-10-11 sensors, the roadside computing unit can perceive information about obstacles and passing vehicles at intersections in real time, and perform corresponding data calculations. The calculation results are broadcast through the roadside unit (RSU).
[0045] The V2X-based road safety device for autonomous driving systems can be applied in practical scenarios and its control over vehicles is described below:
[0046] At the intersection, 12 is a large, long-term obstacle located in the return lane, such as a fan or pumping station. 13-14 are autonomous vehicles located in the outbound and return lanes respectively. Path 15 is the autonomous driving planning path for outbound vehicle 13, and path 16 is the planning path for return vehicle 14.
[0047] At this intersection, due to the influence of large obstacle 12, outbound vehicle 13 and return vehicle 14 will overlap at section A and travel in opposite directions. At this time, due to the obstruction of the alley wall, vehicles 13 and 14 are both in each other's blind spots, and the vehicle's perception sensors cannot achieve beyond-line-of-sight perception.
[0048] The 9-10-11 sensors installed on the roadside can simultaneously detect the speed, heading, and other relevant information of vehicles 13 and 14. Therefore, in this scenario, the roadside unit needs to fuse and calculate the detected vehicle information and broadcast it via the RSU. At this time, autonomous vehicles 13 and 14 can simultaneously receive the information provided by the roadside unit and, based on this information, control acceleration and throttle to perform acceleration, deceleration, or braking operations, ensuring that the two vehicles do not simultaneously reach position A on the road. This significantly improves driving safety, thus achieving road safety settings for a V2X-based autonomous driving system.
[0049] Example 2
[0050] like Figure 4 As shown, a road safety control method for a V2X-based unmanned driving system is characterized by comprising:
[0051] S1 collects environmental data of the intersection area through roadside sensing equipment groups. The environmental data includes lidar data, millimeter-wave radar data and image data.
[0052] S2 utilizes the roadside computing unit (MEC) to fuse and calculate the collected lidar data, millimeter-wave radar data, and image data to obtain global intersection data. This global intersection data includes: the position, speed, and heading of all vehicles within the intersection, as well as the position and size of obstacles in the intersection area, including:
[0053] S21: Processing LiDAR data: Filtering and denoising the LiDAR point cloud, aggregating adjacent point clouds into point cloud clusters using a clustering algorithm; determining the target type based on the positional changes of the point cloud clusters within a continuous scanning cycle, identifying static targets as those whose positions remain unchanged and dynamic targets as those whose positions change; calculating the three-dimensional coordinates of the center point of each point cloud cluster in the roadside coordinate system.
[0054] S22: Processing millimeter-wave radar data: Performing a Fast Fourier Transform on the echo signal received by the millimeter-wave radar to extract the Doppler frequency shift in the spectrum; calculating the target's radial velocity according to the Doppler effect formula, where... The value is the Doppler frequency shift, and v is the target velocity. The transmission frequency is given, θ is the angle between the target's direction of motion and the radar's radial direction, and c is the speed of light; the radial velocity is compared with the three-dimensional coordinates of the corresponding dynamic target in step S21. The matching process is performed to obtain the radial velocity value of each dynamic target;
[0055] S23: Image data processing: Image preprocessing includes denoising and enhancement; target detection is performed on the preprocessed image using a convolutional neural network, outputting target bounding boxes and category labels, including vehicles and obstacles; the same target in consecutive frames is tracked, and the target's motion direction angle is calculated based on the target's displacement vector in the image coordinate system;
[0056] S24: Perform spatiotemporal registration of the target's three-dimensional coordinates obtained in step S21, the dynamic target's radial velocity obtained in step S22, and the target's category and motion direction angle obtained in step S23: unify the data from each sensor to the same time reference, transform the coordinate systems of each sensor to a unified roadside coordinate system; associate the same target data from different sensors using the nearest neighbor matching algorithm to generate comprehensive target data including position, velocity, category, and direction;
[0057] S25: Construct global intersection data based on comprehensive target data: classify each target in the target data: dynamic targets classified as vehicles are assigned to the vehicle target set; static targets classified as obstacles are assigned to the obstacle target set;
[0058] For each vehicle target in the vehicle target set, generate vehicle state information: use the three-dimensional coordinates of the vehicle target as the vehicle position; calculate the vehicle's velocity vector in the roadside coordinate system based on the vehicle target's radial velocity and direction of motion, and use it as the vehicle velocity; use the vehicle target's direction of motion as the vehicle heading.
[0059] For each obstacle target in the obstacle target set, generate obstacle state information: use the three-dimensional coordinates of the obstacle target as the obstacle position; calculate the length, width, and height of the obstacle based on the boundary of the point cloud cluster corresponding to the obstacle target, as the obstacle size;
[0060] All vehicle status information and obstacle status information are organized according to a preset data format to generate structured data containing timestamps, number of targets, ID of each target and its corresponding status information;
[0061] Structured data is output as global intersection data, which includes the position, speed, and heading of all vehicles within the intersection, as well as the position and size of obstacles in the intersection area.
[0062] In this scheme, static targets refer to objects whose position remains unchanged across multiple consecutive scanning cycles of the roadside lidar. In the environment of mine roadway intersections, static targets mainly include: roadway walls, support equipment, fixed ventilation ducts, parked equipment or vehicles, and fallen ore. Although these targets do not move, they may still pose obstacles to passing vehicles and need to be accurately identified and located so that autonomous vehicles can plan avoidance paths.
[0063] Dynamic targets refer to objects whose position changes within a continuous scanning cycle of a roadside lidar. In a mine tunnel environment, dynamic targets mainly refer to unmanned transport vehicles, manned vehicles, and moving work equipment. By determining the dynamic and static attributes of the target, the system can adopt different processing strategies: for dynamic targets, it is necessary to predict their trajectory, while for static targets, it is only necessary to mark the space they occupy.
[0064] Radial velocity refers to the radial velocity component of a target relative to a millimeter-wave radar, i.e., the projection of the target's velocity onto the radar's line-of-sight direction. This is derived from the Doppler effect formula. The calculation yields a value where θ is the angle between the target's actual direction of motion and the radar's radial direction. In a tunnel environment, due to the constraints of the tunnel's orientation, the vehicle's direction of motion is relatively fixed. Therefore, the vehicle's actual speed can be calculated based on the radial velocity and the tunnel's geometric information.
[0065] The motion direction angle refers to the direction of motion of a target in the roadside coordinate system, obtained by analyzing the displacement vector of the target in consecutive image frames. At the intersection of mine roadways, the motion direction angle of a vehicle can clearly indicate which roadway the vehicle came from and which direction it will travel, which is crucial for judging the vehicle's driving intention and predicting its future trajectory. Combined with the target's speed, the vehicle's trajectory at the intersection can be accurately predicted.
[0066] Global intersection data is a comprehensive description of all target information within the entire intersection area provided by the roadside system. It is a structured information set generated by fusing data from three sensors: LiDAR, millimeter-wave radar, and cameras. Based on a unified roadside coordinate system, this data includes the real-time position, velocity vector, and heading angle of all vehicles within the intersection, as well as the position and three-dimensional dimensions of all obstacles. This data is broadcast to vehicles entering the intersection via V2X communication to obtain information about vehicles and obstacles in blind spots obscured by the road walls.
[0067] S3 broadcasts global intersection data via the 5.8GHz frequency band through the roadside communication unit (RSU).
[0068] S4, the driverless vehicle receives global intersection data broadcast by the roadside via the on-board unit (OBU);
[0069] S5, the autonomous vehicle controller fuses the received global intersection data with the vehicle-side environmental data collected by the vehicle-side perception equipment group to obtain fused perception data. This fused perception data includes data within the autonomous vehicle's own perception range and data within its own sensor blind spots, including:
[0070] S51: Process the vehicle-mounted LiDAR data, vehicle-mounted millimeter-wave radar data, and vehicle-mounted camera data collected by the vehicle-mounted perception equipment group, extract the target's position coordinates, velocity vector, and category label, and construct a set of vehicle-mounted perception targets;
[0071] S52: Based on the current position coordinates of the autonomous vehicle and heading angle Based on the detection parameters of the vehicle-mounted sensing equipment group, the sensing blind zone is calculated:
[0072] Acquire the detection parameters of the vehicle-mounted sensing device group, including the maximum detection range of the vehicle-mounted LiDAR. And horizontal field of view α, maximum detection range of vehicle-mounted millimeter-wave radar And detection angle β, effective recognition distance of vehicle camera and field of view angle γ;
[0073] With the current location of the autonomous vehicle With the origin as the starting point, the heading angle is... A local vehicle coordinate system is established using the reference direction. Within this system, the initial sensing area is defined based on the detection parameters of each sensor: Initial sensing area for LiDAR: Initial sensing area of millimeter-wave radar: Initial sensing area of the camera: Where r is the polar radius and φ is the polar angle;
[0074] The initial integrated sensing area is obtained by performing a union operation on the initial sensing areas of the three sensors. ;
[0075] Obtain the location data of the tunnel walls and calculate the occlusion of the tunnel walls on the initial integrated sensing area: represent the tunnel walls as a set of line segments. Each line segment Defined by the coordinates of two endpoints; for the initial integrated sensing area Discretize the boundary to obtain the set of boundary points. For each boundary point Build from vehicle location arrive line segments ; Calculate line segments With all tunnel wall segments The intersection point, if an intersection point exists And the intersection point is located at and Between, the line segment Marked as an obscured area;
[0076] From the initial integrated sensing area By removing all occluded areas, the effective sensing area is obtained. ;Will The boundary is used as the boundary of the vehicle-side perception area, and the local coordinate system of the vehicle is defined. All areas outside of the vehicle's perception area are defined as the vehicle's blind spot.
[0077] S53: Perform coordinate transformation on the global intersection data received in step S4:
[0078] For each target location in the global data of the intersection Based on the location of the driverless vehicle and heading angle The coordinates are then transformed to the vehicle coordinate system using the rotation and translation transformation formula:
[0079] ; Perform the same rotational transformation on the target velocity vector to obtain the set of roadside sensing targets in the vehicle coordinate system;
[0080] S54: Perform data association between the vehicle-side perceived target set and the roadside perceived target set:
[0081] Calculate the position distance for each target i in the vehicle-side perception target set and each target j in the roadside perception target set in the vehicle coordinate system. ;
[0082] Set matching threshold ,when When the target i and target j are selected as candidate matching pairs;
[0083] For each candidate matching pair, calculate the speed similarity: ,in, and These represent the speeds of the two targets, respectively.
[0084] When speed similarity When the value exceeds a preset threshold, target i and target j are confirmed to be the same target. The target is marked as a matched target, and the data of target i perceived by the vehicle is used as the fusion data of the target.
[0085] Traverse the set of roadside sensing targets in the vehicle coordinate system. For unmatched targets: determine whether the target's position is within the defined vehicle-side sensing blind zone; if it is within the blind zone, treat the target as a supplementary target for the blind zone.
[0086] Constructing fused perception data: Merging all target data in the vehicle-side perception target set, as well as roadside perception target data marked as blind spot supplementary targets, to form fused perception data.
[0087] S55: Construct fused sensing data based on data association results:
[0088] Create a list of fusion targets, including target ID, location coordinates, velocity vector, category label, and data source identifier;
[0089] Add all targets in the vehicle-side perception target set to the fusion target list, and set the data source identifier to "vehicle-side";
[0090] Add roadside sensing targets marked as blind spot targets to the fusion target list, and set the data source identifier to "roadside";
[0091] The output contains fusion-aware data that includes a list of fusion targets.
[0092] In this context, a perception blind spot refers to an area that an autonomous vehicle's onboard sensors (LiDAR, millimeter-wave radar, cameras) cannot directly detect due to physical obstructions. In the environment of mine roadway intersections, perception blind spots are primarily caused by roadway walls, especially at T-junctions or crossroads, where vehicles cannot detect oncoming traffic in the transverse roadways. These blind spots are a major cause of intersection collisions. This solution supplements this blind spot information by receiving roadside sensing data, enabling vehicles to perceive targets obscured by roadway walls.
[0093] The maximum detection range, R_laser, represents the farthest distance at which the lidar can effectively detect a target. In a mining environment, this is typically 50–100 meters, but can be reduced by the length of the tunnel and dust levels. The horizontal field of view, α, represents the horizontal scanning range of the lidar, typically 120°–360°. In narrow mine tunnels, even with a 360° field of view, the effective detection range of the lidar is severely limited by the tunnel walls, creating numerous blind spots. These two parameters together determine the fan-shaped sensing area of the lidar under ideal conditions.
[0094] Union operation is a mathematical operation that combines the perception areas of three sensors: LiDAR, millimeter-wave radar, and camera. Because these sensors operate on different principles and are installed in different locations, their perception areas differ: LiDAR offers high accuracy but is susceptible to dust, millimeter-wave radar has strong penetration but lower resolution, and cameras can identify target types but are limited by lighting conditions. Union operation fully leverages the strengths of each sensor to create a maximized integrated perception area. This improves the vehicle's overall perception capabilities in the complex environment of a mine.
[0095] S6, the autonomous vehicle's on-board controller, based on fused perception data, determines whether there is a risk of collision between the vehicle and other vehicles or obstacles in the intersection area, including:
[0096] Extract the location, velocity, category, and data source identifier of all targets from the fused perception data, including vehicle targets and obstacle targets;
[0097] Obtain the current location of the autonomous vehicle. Current speed and heading angle Generate future time series based on a preset time interval Δt. ,in, For each i=1,2,...,n, based on the vehicle kinematics model, predict the autonomous vehicle's speed at each time step. Location The predicted trajectory of this vehicle is obtained. };
[0098] For each vehicle target j in the fused perception data, extract its current position. ,speed and heading angle Based on the future time series T, predict the vehicle target j at each time step. Location The predicted trajectory of vehicle target j is obtained. ;
[0099] For vehicle target j whose data source is identified as a blind spot in vehicle-end perception, the corresponding predicted trajectory is obtained using the geometric boundary data of the roadway. Constraints are applied, including: acquiring the geometric boundary data of the roadway, including the coordinate point sets of the left and right boundary lines of the roadway; and predicting the trajectory of vehicle target j in the vehicle-end perception blind spot. Each predicted position in Calculate the vertical distance from this location to the left and right boundaries of the tunnel. and ; Determine whether the predicted location exceeds the roadway boundary: when or When the predicted position exceeds the drivable area of the roadway, the predicted position is corrected by projecting it perpendicular to the drivable direction onto the nearest roadway boundary. The predicted trajectory of vehicle target j is updated as follows: The positions that do not exceed the boundary remain unchanged.
[0100] Based on the predicted trajectory of this vehicle Calculate whether there is a risk of collision between the autonomous vehicle i and other vehicles or obstacles in the intersection area, including:
[0101] For each autonomous vehicle i, at each moment Calculate the predicted position of this vehicle Euclidean distance between the predicted position of vehicle target j and the predicted position:
[0102] If the data source for vehicle target j is identified as the vehicle-side perception blind spot, the predicted position after constraints is used. Calculate distance ;
[0103] If the data source for vehicle target j is identified as vehicle-side perception, the original predicted location is used. Calculate distance ;
[0104] For each obstacle target k, obtain its position and size information, and construct the boundary rectangle of the obstacle; at each time step Calculate the predicted position of this vehicle Minimum distance to the boundary rectangle of obstacle k ;
[0105] Set a safe distance threshold for vehicle targets. Set a safe distance threshold for obstacle targets. For targets originating from blind spots in vehicle-mounted perception, the corresponding safe distance threshold is increased by a factor of α, where α > 1.
[0106] When any moment exists , making or or or At that time, it was determined that there was a risk of collision;
[0107] Output the collision risk assessment results, including whether there is a collision risk, the minimum distance value, and the corresponding target identifier.
[0108] S7: When a collision risk is detected, the autonomous vehicle controller controls the autonomous vehicle to accelerate or decelerate in order to avoid a collision.
[0109] The foregoing illustrative description of the present application and its embodiments is not restrictive and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. The accompanying drawings are only one embodiment of the present application, and the actual structure is not limited thereto. Therefore, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the present application, such designs should fall within the scope of protection of this application. Furthermore, the word "comprising" does not exclude other elements or steps, and the word "a" preceding an element does not exclude the inclusion of "a plurality" of that element. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.
Claims
1. A road safety control method for a V2X-based unmanned driving system, characterized in that, include: S1 collects environmental data of the intersection area through roadside sensing equipment groups. The environmental data includes lidar data, millimeter-wave radar data and image data. S2 utilizes the roadside computing unit MEC to fuse and calculate the collected lidar data, millimeter-wave radar data, and image data to obtain global intersection data; Global intersection data includes: the position, speed, and heading of all vehicles within the intersection, as well as the position and size of obstacles in the intersection area; S3 broadcasts global intersection data via the 5.8GHz frequency band through the roadside communication unit (RSU). S4, the driverless vehicle receives global intersection data broadcast by the roadside via the on-board unit (OBU); S5, the autonomous vehicle controller fuses the received global intersection data with the vehicle-side environmental data collected by the vehicle-side perception device group to obtain fused perception data; among which, the fused perception data includes data within the autonomous vehicle's own perception range, as well as data within the blind spots of its own sensors; S6: The autonomous vehicle controller uses fused perception data to determine whether there is a risk of collision between the vehicle and other vehicles or obstacles in the intersection area. S7, when a collision risk is detected, the autonomous vehicle controller controls the autonomous vehicle to accelerate or decelerate in order to avoid a collision; S5, obtain fused sensing data, including: Based on the vehicle-side environmental data collected by the vehicle-side sensing equipment group, the location coordinates of the vehicle-side targets are extracted. Based on speed and category, construct a set of vehicle-side perception targets; Based on the current location coordinates of the autonomous vehicle and heading angle Based on the parameters of the vehicle-side sensing equipment group, the boundary of the vehicle-side sensing area is determined, and the area outside the boundary of the vehicle-side sensing area is defined as the vehicle-side sensing blind zone. The global data of the intersection is transformed from the roadside coordinate system to the vehicle coordinate system centered on the autonomous vehicle, resulting in a set of roadside perception targets in the vehicle coordinate system. Target matching is performed on the vehicle-end perception target set and the roadside perception target set in the vehicle coordinate system to obtain fused perception data. The fused perception data includes: targets in the vehicle-end perception target set and roadside perception targets located in the vehicle-end perception blind spot. Define the boundaries of the vehicle-side perception area, and define the area outside the vehicle-side perception area boundaries as the vehicle-side perception blind zone, including: Acquire the detection parameters of the vehicle-mounted sensing equipment group, including the maximum detection range and field of view of the vehicle-mounted lidar, the maximum detection range and detection angle of the vehicle-mounted millimeter-wave radar, and the effective recognition range and field of view of the vehicle-mounted camera; Based on the detection parameters of each sensor, the sensing area of each sensor is determined, and a union operation is performed to obtain the comprehensive sensing area. Obtain the location of the tunnel wall, calculate the occlusion of the tunnel wall on the integrated sensing area, and remove the occluded area from the integrated sensing area; The boundary of the integrated perception area after removing obstructions is taken as the boundary of the vehicle-side perception area. Target matching is performed on the vehicle-side perceived target set and the roadside perceived target set in the vehicle coordinate system to obtain fused perception data, including: Spatial position correlation is performed between the vehicle-side perceived target set and the roadside perceived target set in the vehicle coordinate system, and the positional distance between the targets is calculated. Candidate matching target pairs are determined based on location distance and a preset matching threshold; The motion feature similarity of candidate matching target pairs is verified. Target pairs that meet the similarity conditions are identified as the same target, and vehicle-side perception data is used as the fusion data for matching targets. Determine whether the unmatched targets in the roadside perception target set are located within the determined vehicle-side perception blind zone, and mark the targets located within the blind zone as blind zone supplementary targets; The vehicle-side perception target set and blind spot supplementary targets are merged to generate fused perception data that includes targets within the vehicle's perception range and targets within the perception blind spot.
2. The road safety control method for a V2X-based unmanned driving system according to claim 1, characterized in that: S2, obtains the global intersection data, including: The lidar data is filtered and point cloud clusters are extracted using a clustering algorithm. Static and dynamic targets are distinguished based on the point cloud clusters, and the three-dimensional coordinates of each target in the roadside coordinate system are calculated. Perform a fast Fourier transform on the millimeter-wave radar data to calculate the radial velocity of each dynamic target; Based on the image data, the target category and direction of motion are obtained through a target detection algorithm. The target categories include vehicles and obstacles. The target's three-dimensional coordinates, dynamic target radial velocity, target category, and direction of motion are correlated to obtain target data; Based on the target data, construct global intersection data.
3. The road safety control method for a V2X-based unmanned driving system according to claim 2, characterized in that: Based on the target data, construct global intersection data, including: Based on the target category, the target data is divided into a vehicle target set and an obstacle target set; For each vehicle target in the vehicle target set, generate vehicle status data: Use the three-dimensional coordinates of the vehicle target as the vehicle position; Based on the radial velocity and direction of motion of the vehicle target, calculate the velocity vector of the corresponding vehicle target in the roadside coordinate system, which is used as the vehicle velocity; Use the direction of the vehicle target's movement as the vehicle's heading; For each obstacle in the obstacle target set, generate obstacle state data: Use the three-dimensional coordinates of the obstacle target as the obstacle's position; Based on the point cloud cluster boundary corresponding to the obstacle target, calculate the length, width, and height of the obstacle as the obstacle size; Vehicle status data and obstacle status data are output as global data for the intersection.
4. The road safety control method for a V2X-based unmanned driving system according to claim 1, characterized in that: S6, the autonomous vehicle's on-board controller, based on fused perception data, determines whether there is a risk of collision between the vehicle and other vehicles or obstacles in the intersection area, including: Extract the location, velocity, category, and data source identifier of all targets from the fused perception data, including vehicle targets and obstacle targets; Obtain the current location of the autonomous vehicle. Current speed and heading angle According to the preset time interval Generating future time series ,in, Based on the vehicle kinematics model, predict the autonomous vehicle at each moment. Location The predicted trajectory of this vehicle is obtained. ; For each vehicle target j in the fused perception data, extract its current position. ,speed and heading angle Based on the future time series T, predict the vehicle target j at each time step. Location The predicted trajectory of vehicle target j is obtained. ; For vehicle target j whose data source is identified as a blind spot in vehicle-end perception, the corresponding predicted trajectory is obtained using the geometric boundary data of the roadway. Apply constraints; Based on the predicted trajectory of this vehicle Calculate whether there is a risk of collision between the autonomous vehicle i and other vehicles or obstacles in the intersection area.
5. The road safety control method for a V2X-based unmanned driving system according to claim 4, characterized in that: Using the geometric boundary data of the tunnel to predict the corresponding trajectory Constraints include: Obtain the geometric boundary data of the tunnel to determine the drivable area of the tunnel; Predicted trajectory of vehicle target j in the blind spot of vehicle-side perception Whether each predicted location is within the drivable area of the roadway; For predicted positions that are outside the drivable area, the position is corrected according to the roadway boundary constraints so that the corrected position is within the drivable area, thus obtaining the constrained predicted trajectory.
6. The road safety control method for a V2X-based unmanned driving system according to claim 4, characterized in that: Calculate whether there is a risk of collision between the autonomous vehicle i and other vehicles or obstacles in the intersection area, including: For each autonomous vehicle i, at each moment Calculate the predicted position of this vehicle Euclidean distance between the predicted position of vehicle target j and the predicted position: If the data source for vehicle target j is identified as the vehicle-side perception blind spot, the predicted position after constraints is used. Calculate distance ; If the data source for vehicle target j is identified as vehicle-side perception, the original predicted location is used. Calculate distance ; For each obstacle target k, its position and size information are obtained, and the boundary rectangle of the obstacle is constructed; at each time ti, the predicted position of the vehicle is calculated. Minimum distance to the boundary rectangle of obstacle k ; Set a safe distance threshold for vehicle targets. Set a safe distance threshold for obstacle targets. For targets originating from blind spots in vehicle-mounted perception, the corresponding safe distance threshold is increased by a factor of α, where α >
1. When any moment exists , making or At that time, it was determined that there was a risk of collision; Output the collision risk assessment results, including whether there is a collision risk, the minimum distance value, and the corresponding target identifier.
7. A road safety device for a V2X-based unmanned driving system, used to perform the method according to any one of claims 1 to 6, characterized in that, include: Roadside systems, installed at road intersections, include: The roadside computing unit (MEC) performs fusion calculations on the data collected by roadside sensing devices. The roadside communication unit (RSU) broadcasts the fusion calculation results via a 5.8 GHz frequency band signal and receives signals in the same frequency band from the vehicle-mounted communication equipment. The roadside sensing equipment group includes lidar, millimeter-wave radar and cameras installed on the roadside, used to sense obstacles and passing vehicle data in the intersection area in real time; The vehicle-mounted system, installed on the autonomous vehicle, includes: The on-board unit (OBU) broadcasts vehicle data and receives broadcast data from the roadside unit (RSU). The autonomous vehicle controller receives environmental data from the vehicle-mounted sensing device group and roadside data received by the on-board communication unit (OBU), performs fusion calculations, and then controls the vehicle's steering and speed. The vehicle-mounted sensing equipment group includes lidar, millimeter-wave radar, and cameras installed on the autonomous vehicle; The roadside system acquires global data of the intersection area through perception fusion and transmits the global data to the autonomous vehicles entering the intersection area through V2X communication, so that the autonomous vehicles can obtain beyond-line-of-sight perception capabilities.
Citation Information
Patent Citations
Cooperative intelligent driving system based on V2X roadside message
CN118337813A
Roadside sensing equipment
CN220323539U
