Multi-source information fusion method for decision-making on driving behavior of unmanned underground transport vehicles
By dividing the perception information of the underground unmanned assisted transport vehicle into multiple levels and setting priorities, and designing multiple driving modes, the problem of vehicle behavior decision-making in the underground environment is solved, realizing intelligent driving and safety assurance.
Patent Information
- Application Number
- CN202311660605.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-05
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-12-05
AI Technical Summary
Existing technologies struggle to effectively utilize multi-source sensing information for vehicle behavior decisions in underground environments, especially in coal mines where lighting conditions vary greatly, dusty and foggy sections are common, and tunnels are uneven. This makes it impossible to directly apply ground-based unmanned vehicle driving behavior decision-making methods, and small underground accidents can escalate into major accidents.
By employing a multi-source information fusion method, the perception information of the underground unmanned assisted transport vehicle is divided into an early warning layer, a target layer, and a road condition layer. Priorities are set according to the degree of risk, and conventional driving, collision avoidance driving, and obstacle avoidance driving modes are designed. Autonomous decision-making is carried out in combination with braking distance and collision time coefficient.
It enables intelligent driving of unmanned assisted transport vehicles underground, improves the stability, intelligence and safety of the system, provides early warning functions, and ensures safe vehicle operation.
Smart Images

Figure CN117719559B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of digital production and unmanned driving technology in coal mines, and in particular to a decision-making method for driving behavior of unmanned underground assisted transport vehicles based on multi-source information fusion. Background Technology
[0002] Digitalization, intelligentization, and unmanned operation are inevitable trends in the further development of coal mines. Unmanned auxiliary transport vehicles are inextricably linked to underground production efficiency, making them a primary focus of research and implementation for intelligent underground equipment.
[0003] Driving behavior decision-making technology for unmanned ground vehicles is currently a hot research topic. Unmanned vehicles make driving behavior decisions by collecting sensor information and combining it with autonomous decision-making methods. Patent authorization number CN113759390B proposes an autonomous driving method for mining trucks based on virtual lidar technology. This method uses a single lidar and IMU combined with virtual radar to obtain perception information to assist autonomous driving. However, the perception capability of a single sensor in the coal mine environment is poor, making it difficult to guarantee the safe completion of autonomous driving tasks. Patent publication numbers CN116552552A and CN116481833A, among others, disclose driving behavior decision-making methods for unmanned vehicles based on various advanced technologies. However, the above-mentioned prior art only describes driving behavior decision-making methods for unmanned vehicles in ground environments, and has the following shortcomings for underground environments:
[0004] 1. Coal mine environments include dusty and foggy sections, and the lighting conditions vary greatly. Vehicle operation is affected by changes in roadways, so surface vehicle environmental perception technology and driving behavior decision-making are difficult to use directly in underground environments.
[0005] 2. Compared with the multi-source sensing information on the surface, the sensing information underground is more diverse, including sensing information that is rarely encountered on the surface, such as tunnel morphology, water puddle size, and turnouts. Moreover, underground tunnels are often uneven, and the sensing information is diverse and abundant. How to make reasonable use of the large amount of multi-source sensing information has become a problem.
[0006] 3. Minor accidents underground can escalate into major accidents. Therefore, drivers should prioritize safety over comfort when driving vehicles. Summary of the Invention
[0007] Purpose of the Invention: To address the above-mentioned problems, the purpose of this invention is to provide a multi-source information fusion method for decision-making on driving behavior of unmanned underground auxiliary transport vehicles. This method addresses the complexities of coal mine environments, such as dusty and foggy roads, significant variations in lighting, and the impact of roadway changes on vehicle operation. By fusing multi-source sensor information and classifying the information according to risk levels, the method facilitates further vehicle behavior decision-making, thereby enabling intelligent driving of underground auxiliary transport vehicles. Based on existing and predictive information, the unmanned auxiliary transport vehicle can further implement early warning functions, ensuring the vehicle's safe operation and effectively improving the stability, intelligence, and safety of the unmanned driving system for underground auxiliary transport vehicles.
[0008] Technical solution: One aspect of the present invention provides a method for decision-making on driving behavior of an unmanned underground assisted transport vehicle based on multi-source information fusion, comprising:
[0009] Utilize onboard sensors to obtain multi-source sensing information of the vehicle while it is traveling underground;
[0010] Multi-source perception information is divided into levels according to the degree of risk, and different levels of perception information are prioritized according to the degree of threat to vehicle safe driving.
[0011] Driving modes are designed based on operating conditions;
[0012] The acquired perception information is transformed into unmanned assisted transport vehicle behavior information according to different levels. Based on the unmanned assisted transport vehicle behavior information and safety assessment standards, the vehicle autonomously decides on the driving mode and realizes the autonomous driving mode switching of the vehicle.
[0013] Furthermore, multi-source perception information is obtained by utilizing vehicle-mounted sensors during the vehicle's underground journey. This multi-source perception information is then categorized into levels based on risk level, including:
[0014] Multi-source sensing information is divided into three levels: early warning layer, target layer, and road condition layer.
[0015] Among them, the early warning layer perception information is risk warning information that is autonomously judged based on the information obtained from existing vehicle-mounted sensors, including pedestrian behavior warning information, following vehicle behavior warning information, and alleyway deformation warning information;
[0016] The target layer perception information is the obstacle perception and recognition information in front of the vehicle, including obstacle type recognition information and obstacle depth perception information;
[0017] The road condition perception information is the road condition perception and identification information in the working environment of the vehicle, including the identification information of the type of damage to the tunnel wall, the perception information of the degree of damage to the tunnel wall, the identification of water seepage and water accumulation in the floor, the perception information of track anomalies, and the identification information of turnout anomalies.
[0018] Furthermore, prioritizing different levels of perceived information based on the degree of threat to vehicle safe driving includes:
[0019] Road condition perception information determines whether a vehicle can pass through the current lane. It is the basic perception information for autonomous driving and is set as the first priority.
[0020] The target layer perception information refers to obstacles in front of the unmanned assisted transport vehicle that affect the vehicle's driving safety. The unmanned assisted transport vehicle uses the perception information of this layer to plan obstacle avoidance trajectory and make other behavioral decisions, and sets it as the second priority.
[0021] The early warning layer's perception information is based on long-term records or sensor decision-making information currently in the warning area. This information indicates that there is a certain risk but passage is possible, and it is set as the third priority.
[0022] Furthermore, the driving modes designed according to operating conditions include:
[0023] Based on the three layers of multi-source perception information and their priority relationships, three driving modes are designed according to different operating conditions: normal driving mode, collision avoidance driving mode, and obstacle avoidance driving mode.
[0024] Among them, the normal driving mode corresponds to the warning layer perception information, the collision avoidance driving mode corresponds to the target layer perception information, and the obstacle avoidance driving mode corresponds to the warning layer perception information, the target layer perception information, and the road condition layer perception information.
[0025] Furthermore, the safety assessment criteria include: braking distance and collision time coefficients introduced by the driving behavior decision-making of unmanned assisted transport vehicles, wherein the calculation expression for braking distance D is:
[0026]
[0027] In the formula, t1 and t2 represent the response time of robot perception and decision-making, respectively; vcc and voc represent the current speed of the unmanned assisted transport vehicle and the longitudinal speed of the obstacle in front, respectively; voc = 0 indicates that the obstacle is static; voc ≠ 0 indicates that the obstacle is dynamic; and acc is the relative acceleration between the vehicle and the obstacle in front.
[0028] Collision time coefficient tcc -1 The calculation expression is:
[0029]
[0030] In the formula, dov represents the current distance between the obstacle and the vehicle, and L represents the length of the obstacle.
[0031] Furthermore, the autonomous decision-making process for driving modes based on the behavior information and safety assessment standards of the unmanned assisted transport vehicle includes the following steps:
[0032] S21. If the obstacle that suddenly appears in the perception field of the unmanned assisted transport vehicle is a sudden obstacle, then proceed to the safety assessment and use the safety assessment criteria to make driving behavior decisions; otherwise, proceed to step S22.
[0033] S22, when the obstacle is not a sudden obstacle, the vehicle makes driving behavior decisions based on whether the distance between the obstacle and the front of the vehicle reaches the collision avoidance distance or obstacle avoidance distance.
[0034] Furthermore, step S21 specifically includes the following process:
[0035] Determine whether a sudden obstacle has speed and its direction. If there is no speed, the obstacle is a static obstacle. Calculate the minimum braking distance and compare it with the depth of field of the detected obstacle. Based on the comparison result, switch driving modes autonomously.
[0036] When obstacles have both speed and acceleration, the collision time coefficient is used as the evaluation standard, and the driving mode is automatically switched based on the comparison results.
[0037] Furthermore, step S22 specifically includes the following process:
[0038] When the onboard sensors detect that the distance between the obstacle and the unmanned assisted transport vehicle is greater than the collision avoidance distance or there is no obstacle at present, the unmanned assisted transport vehicle is in normal driving mode;
[0039] When the onboard sensors detect that the distance between the obstacle and the unmanned assisted transport vehicle is less than the collision avoidance distance, the unmanned assisted transport vehicle autonomously switches to the collision avoidance driving mode.
[0040] When the vehicle's onboard sensors detect an obstacle within a safe distance and before it reaches the obstacle avoidance distance, the vehicle slows down.
[0041] When the onboard sensors detect an obstacle within the obstacle avoidance distance, the unmanned assisted transport vehicle autonomously switches to obstacle avoidance driving mode.
[0042] Furthermore, when the onboard sensors detect an obstacle within the obstacle avoidance distance, the unmanned assisted transport vehicle autonomously switches to obstacle avoidance driving mode, including:
[0043] When the vehicle collects road condition perception information, and this causes the height and width of the alley to be lower than the height and width of the vehicle after the alley converges, the unmanned assisted transport vehicle will stop urgently.
[0044] When the vehicle is a rail vehicle, it can only travel under the constraints of the track. At this time, the vehicle will make an emergency stop until the obstacle is far away, and then it will automatically switch to the normal driving mode or the collision avoidance driving mode according to the distance.
[0045] When the vehicle is an unmanned assisted transport vehicle, the optimal obstacle avoidance path is obtained based on local path planning.
[0046] Another aspect of the present invention provides a multi-source information fusion-based decision-making system for the driving behavior of an unmanned assisted transport vehicle in underground mines, the system comprising a perception module, a decision-making module, and a control module;
[0047] The perception module includes a variety of vehicle-mounted sensors, which are used to perceive and identify moving and static, near and far multimodal obstacles and road conditions to obtain multi-source perception information.
[0048] The decision-making module is used to make decisions based on multi-source sensing information;
[0049] The control module is used to control the underlying hardware devices to complete autonomous driving based on decision information.
[0050] Beneficial effects: Compared with the prior art, the significant advantages of this invention are:
[0051] 1. This invention proposes an information fusion strategy for managing and rationally using a large amount of multi-source information. Based on the different impacts of the underground environment on the risk of vehicle driving, the multi-source information is divided into three priority levels to achieve multi-source information fusion.
[0052] 2. Based on multi-source fusion data, this invention proposes three driving behavior modes for unmanned assisted transport vehicles: conventional driving mode, collision avoidance driving mode, and obstacle avoidance driving mode. These three modes correspond to different levels of multi-source information. The vehicle can autonomously switch driving modes based on the perceived information to achieve safe and reasonable intelligent behavior decisions.
[0053] 3. This invention proposes a driving behavior decision-making architecture for unmanned assisted transport vehicles specifically designed for underground environments, which solves the problem that driving behavior decision-making methods for ground-based unmanned vehicles cannot be directly applied to unmanned assisted transport vehicles underground due to the special characteristics of the underground coal mine environment. Attached Figure Description
[0054] Figure 1 This is a schematic diagram of the behavior decision-making architecture of the trackless unmanned assisted transport vehicle in the embodiment.
[0055] Figure 2 This is a schematic diagram of the behavior decision-making architecture of the underground rail-guided unmanned assisted transport vehicle in the embodiment.
[0056] Figure 3 This is a schematic diagram illustrating how the unmanned underground assisted transport vehicle decides its driving mode based on perception fusion data in the embodiment.
[0057] Figure 4 This is a decision-making flowchart for the unmanned underground assisted transport vehicle in the embodiment;
[0058] Figure 5 This is a schematic diagram illustrating the fusion of information from multiple sensors in the embodiment;
[0059] Figure 6 This is a diagram showing the relationship between the entire system of the unmanned underground assisted transport vehicle, its various sensors, and decision-making modules in the embodiment. Detailed Implementation
[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.
[0061] Example 1
[0062] The multi-source information fusion-based decision-making method for driving behavior of unmanned underground transport vehicles described in this embodiment includes:
[0063] Utilize onboard sensors to obtain multi-source sensing information of the vehicle while it is traveling underground;
[0064] Multi-source perception information is divided into levels according to the degree of risk, and different levels of perception information are prioritized according to the degree of threat to vehicle safe driving.
[0065] Driving modes are designed based on operating conditions;
[0066] The acquired perception information is transformed into unmanned assisted transport vehicle behavior information according to different levels. Based on the unmanned assisted transport vehicle behavior information and safety assessment standards, the vehicle autonomously decides on the driving mode and realizes the autonomous driving mode switching of the vehicle.
[0067] In one example, multi-source sensing information is obtained using onboard sensors while the vehicle is traveling underground. This multi-source sensing information is then categorized into levels based on risk level, including:
[0068] Multi-source sensing information is divided into three levels: early warning layer, target layer, and road condition layer. The early warning layer provides risk warnings based on information acquired from existing onboard sensors, including pedestrian behavior warnings, following behavior warnings, and tunnel deformation warnings. The target layer provides obstacle perception and identification information ahead of the vehicle, including obstacle type identification and obstacle depth perception; obstacles include pedestrians, mining cars, sandstone, etc. The road condition layer provides road condition perception and identification information within the vehicle's working environment, including tunnel wall damage type identification, tunnel wall damage degree identification, floor seepage and water accumulation identification, track anomaly identification, and switch anomaly identification. Tunnel wall damage includes roof falls, spalling, floor bulging, and bulging of the sides; track anomalies refer to detachment and deformation.
[0069] like Figure 3 As shown, prioritizing different levels of perception information based on the degree of threat to vehicle safe driving includes:
[0070] Road condition perception information determines whether a vehicle can pass through the current lane. It is the basic perception information for autonomous driving and is set as the first priority.
[0071] The target layer perception information refers to obstacles in front of the unmanned assisted transport vehicle that affect the vehicle's driving safety. The unmanned assisted transport vehicle uses the perception information of this layer to plan obstacle avoidance trajectory and make other behavioral decisions, and sets it as the second priority.
[0072] The early warning layer's perception information is based on long-term records or sensor decision-making information currently in the warning area. This information indicates that there is a certain risk but passage is possible, and it is set as the third priority.
[0073] When unmanned assisted transport vehicles operate autonomously underground, they acquire a large amount of multi-source perception information from onboard sensors, making rapid autonomous decision-making impossible. Therefore, hierarchical classification of this multi-source information is necessary. However, simple classification only facilitates the processing of multi-source perception information and cannot directly optimize the accuracy of driving behavior decisions. Therefore, it is necessary to prioritize three risk levels based on common safety and obstacle avoidance issues encountered in underground vehicle driving to address different levels of decision-making.
[0074] In one example, such as Figure 1-3 As shown, the driving modes designed according to operating conditions include:
[0075] Based on the three layers of multi-source perception information and their priority relationships, three driving modes are designed according to different operating conditions: normal driving mode, collision avoidance driving mode, and obstacle avoidance driving mode.
[0076] Among them, the normal driving mode corresponds to the warning layer perception information, the collision avoidance driving mode corresponds to the target layer perception information, and the obstacle avoidance driving mode corresponds to the warning layer perception information, the target layer perception information, and the road condition layer perception information.
[0077] Specifically, the safety assessment criteria include: braking distance and collision time coefficient introduced by the driving behavior decision-making of unmanned assisted transport vehicles, wherein the calculation expression for braking distance D is:
[0078]
[0079] In the formula, t1 and t2 represent the robot's perception and decision-making response times, respectively; vcc and voc represent the current speed of the unmanned assisted transport vehicle and the longitudinal velocity of the obstacle ahead, respectively; voc = 0 indicates the obstacle is static, and voc ≠ 0 indicates the obstacle is dynamic; acc is the relative acceleration between the vehicle and the obstacle ahead. When encountering an obstacle, the above braking distance can be used as a safety assessment standard to determine whether the obstacle poses a direct threat, thus enabling driving behavior decisions such as emergency braking. The unmanned assisted transport vehicle's driving behavior decision-making incorporates braking distance and collision time coefficients to cope with sudden and dynamic obstacles.
[0080] To better address obstacles that appear suddenly or are relatively close to the robot, a collision time coefficient (tcc) is introduced as another evaluation criterion. -1 The calculation expression is:
[0081]
[0082] In the formula, dov represents the current distance between the obstacle and the vehicle, and L represents the length of the obstacle. This safety assessment standard helps determine the time of collision at the current vehicle speed and obstacle speed.
[0083] like Figure 4 As shown, the autonomous driving mode decision based on the unmanned assisted transport vehicle's behavior information and safety assessment standards includes the following steps:
[0084] S21. If the obstacle that suddenly appears in the perception field of the unmanned assisted transport vehicle is a sudden obstacle, then proceed to the safety assessment and use the safety assessment criteria to make driving behavior decisions; otherwise, proceed to step S22.
[0085] S22, when the obstacle is not a sudden obstacle, the vehicle makes driving behavior decisions based on whether the distance between the obstacle and the front of the vehicle reaches the collision avoidance distance or obstacle avoidance distance.
[0086] Specifically, step S21 includes the following process:
[0087] Determine whether a sudden obstacle has speed and its direction. If there is no speed, the obstacle is a static obstacle. Calculate the minimum braking distance and compare it with the depth of field of the detected obstacle. Based on the comparison result, switch driving modes autonomously.
[0088] When obstacles have both speed and acceleration, the collision time coefficient is used as the evaluation standard, and the driving mode is automatically switched based on the comparison results.
[0089] In one example, such as Figure 4 As shown, the collision warning threshold The braking threshold is set to 0.3. is set to 0.5. In step 21, when the obstacle is a static obstacle, the relative magnitudes of the braking distance D and the anti-collision distance dca are compared. When D≥dca, the vehicle switches to the obstacle-avoidance driving mode, and it is determined whether the obstacle is a road condition layer obstacle. If yes, the vehicle performs an emergency stop behavior at this time, and until the obstacle moves away, it automatically switches to the normal driving mode or the anti-collision driving mode according to the distance;
[0090] When D<dca, the relative magnitudes of the braking distance D and the obstacle-avoidance distance doa are compared. If D≥doa, the vehicle switches to the anti-collision driving mode; if D<doa, the vehicle switches to the normal driving mode.
[0091] In step 21, when the obstacle is a dynamic obstacle, the time-to-collision coefficient and the collision warning threshold are compared braking threshold in relative magnitude. When 0<1 / tcc≤tcc min , the vehicle switches to the normal driving mode; when tcc min <1 / tcc≤tcc max , the vehicle switches to the anti-collision driving mode. Otherwise, it is determined whether the obstacle has moved away. If the obstacle has not moved away, the vehicle performs an emergency stop, sends a report, and waits for the staff to clear the obstacle; if the obstacle has moved away, the vehicle switches to the normal driving mode. When the destination is reached, the driving behavior decision is ended.
[0092] As Figure 4 shown, step S22 specifically includes the following process:
[0093] When the on-board sensor detects that the distance from the obstacle to the unmanned assisted transport vehicle is greater than the anti-collision distance or there is no obstacle at present, the unmanned assisted transport vehicle is in the normal driving mode;
[0094] When the on-board sensor detects that the distance from the obstacle to the unmanned assisted transport vehicle is less than the anti-collision distance, the unmanned assisted transport vehicle automatically switches to the anti-collision driving mode;
[0095] When the on-board sensor detects that the obstacle is within the safe distance but not within the obstacle-avoidance distance, the vehicle travels at a reduced speed;
[0096] When the on-board sensor detects that the obstacle is within the obstacle-avoidance distance, the unmanned assisted transport vehicle automatically switches to the obstacle-avoidance driving mode.
[0097] The obstacle avoidance driving mode corresponds to the warning layer, target layer, and road condition layer. At this time, the vehicle is in a high-risk area, highly susceptible to accidents. The vehicle's decision-making is based primarily on road condition perception information within the vehicle's working environment, including information on the type of road wall damage (roof fall, sidewall collapse, floor bulge, etc.), the degree of road wall damage (roof fall, sidewall collapse, floor bulge, etc.), water seepage and accumulation in the floor, track anomalies (detachment, deformation), and switch anomalies. It also relies on obstacle perception information such as the type of obstacle (pedestrians, mine cars, sandstone, etc.) and the depth perception information of obstacles (pedestrians, mine cars, sandstone, etc.) in front of the vehicle. When the onboard sensors detect an obstacle within the obstacle avoidance distance, the unmanned assisted transport vehicle autonomously switches to obstacle avoidance driving mode, including:
[0098] When the vehicle collects road condition perception information, and this causes the height and width of the alley to be lower than the height and width of the vehicle after the alley converges, the unmanned assisted transport vehicle will stop urgently.
[0099] The aforementioned road condition perception information includes: abnormal track conditions (detachment, deformation), abnormal switch conditions, or damage to the roadway walls (roof collapse, sidewall spalling, floor bulging, sidewall bulging, etc.) in front of the vehicle.
[0100] When the vehicle is a rail vehicle, it can only travel under the constraints of the track. At this time, the vehicle will make an emergency stop until the obstacle is far away, and then it will automatically switch to the normal driving mode or the collision avoidance driving mode according to the distance.
[0101] When the vehicle is an unmanned assisted transport vehicle, the optimal obstacle avoidance path is obtained based on local path planning.
[0102] This invention discloses a multi-source information fusion method for decision-making in the driving behavior of unmanned underground assisted transport vehicles. It categorizes multi-source perception information collected from various sensors into different levels based on driving risk and assigns priorities. Decision-making methods are then specified according to the impact of each level on driving safety, thus constructing a vehicle driving behavior decision-making architecture. The specific steps are: obtaining multi-source information about the underground environment from multiple sensors; classifying the information into three levels according to risk severity to achieve multi-source information fusion; and designing decision-making strategies for the vehicle based on different levels to convert the acquired perception information into vehicle behavior information. This invention integrates multi-source sensor information and classifies it according to risk levels to achieve information fusion, facilitating further vehicle behavior decision-making and enabling intelligent driving of underground assisted transport vehicles. Based on existing perception and prediction information, the unmanned assisted transport vehicle can further implement early warning functions, providing assurance for the vehicle's safe operation. This effectively improves the stability, intelligence, and safety of the unmanned driving system for underground assisted transport vehicles, and is of great significance for the implementation of unmanned driving technology in coal mines.
[0103] Example 2
[0104] This embodiment proposes a multi-source information fusion-based driving behavior decision-making system for underground unmanned assisted transport vehicles, which is used in the multi-source information fusion-based driving behavior decision-making method for underground unmanned assisted transport vehicles described in Embodiment 1 above. The decision-making system includes a perception module, a decision-making module, and a control module.
[0105] The perception module includes a variety of vehicle-mounted sensors, which are used to perceive and identify moving and static, near and far multimodal obstacles and road conditions to obtain multi-source perception information.
[0106] The decision-making module is used to make decisions based on multi-source sensing information;
[0107] The control module is used to control the underlying hardware devices to complete autonomous driving based on decision information.
[0108] The underground environment is complex, characterized by limited space, insufficient lighting, and fluctuating dust and fog. Different sensing sensors have specialized characteristics due to their different working principles, making it difficult to achieve autonomous driving for unmanned assisted transport vehicles underground using a single vehicle-mounted sensing sensor. Therefore, such as Figure 5-6 As shown, the unmanned underground transport vehicle system designed in this embodiment includes a perception module, a decision-making module, and a control module. The perception module is equipped with various sensors, including lidar, millimeter-wave radar, ultrasonic radar, and cameras. The decision-making module makes decisions based on multi-source perception information. The control module controls the underlying hardware equipment to complete unmanned driving based on the decision information. Multiple sensors can perceive and identify multi-modal obstacles and road conditions, including moving and stationary obstacles, and obstacles at different distances, obtaining multi-source perception information.
Claims
1. A method for driving behavior decision of an unmanned auxiliary transport vehicle in a mine based on multi-source information fusion, characterized in that, include: Utilize onboard sensors to obtain multi-source sensing information of the vehicle while it is traveling underground; Multi-source perception information is categorized into levels based on risk level, and these levels are prioritized based on the degree of threat to vehicle safe driving; specifically: Multi-source sensing information is divided into three levels: early warning layer, target layer, and road condition layer. Among them, the early warning layer perception information is risk warning information that is autonomously judged based on the information obtained from existing vehicle-mounted sensors, including pedestrian behavior warning information, following vehicle behavior warning information, and alleyway deformation warning information; The target layer perception information is the obstacle perception and recognition information in front of the vehicle, including obstacle type recognition information and obstacle depth perception information; The road condition perception information is the road condition perception and identification information in the working environment of the vehicle, including the identification information of the type of damage to the tunnel wall, the perception information of the degree of damage to the tunnel wall, the identification of water seepage and water accumulation in the floor, the perception information of track anomalies, and the identification information of turnout anomalies. Driving modes are designed based on operating conditions; specifically: Based on the three layers of multi-source perception information and their priority relationships, three driving modes are designed according to different operating conditions: normal driving mode, collision avoidance driving mode, and obstacle avoidance driving mode. Among them, the normal driving mode corresponds to the warning layer perception information, the collision avoidance driving mode corresponds to the target layer perception information, and the obstacle avoidance driving mode corresponds to the warning layer perception information, the target layer perception information and the road condition layer perception information. The obtained perception information is transformed into unmanned assisted transport vehicle behavior information according to different levels. Based on the unmanned assisted transport vehicle behavior information and safety assessment standards, the vehicle autonomously decides the driving mode and realizes the autonomous driving mode switching of the vehicle. The behavioral information of the unmanned assisted transport vehicle includes braking distance and collision time coefficient, where the formula for calculating braking distance D is: ; In the formula, and These represent the response times of the robot's perception and decision-making, respectively. and These represent the current speed of the unmanned transport vehicle and the longitudinal speed of the obstacle ahead, respectively. This indicates that the obstacle is static. This indicates that the obstacle is dynamic. It is the relative acceleration between the vehicle and the obstacle in front; Collision time coefficient The calculation expression is: ; In the formula, Indicates the current distance between the obstacle and the vehicle. Indicates the length of the obstacle.
2. The method for decision-making on driving behavior of unmanned underground assisted transport vehicles based on multi-source information fusion according to claim 1, characterized in that, Prioritizing different levels of perceived information based on the degree of threat to vehicle safe driving includes: Road condition perception information determines whether a vehicle can pass through the current lane. It is the basic perception information for autonomous driving and is set as the first priority. The target layer perception information refers to obstacles in front of the unmanned assisted transport vehicle that affect the vehicle's driving safety. The unmanned assisted transport vehicle uses the perception information of this layer to plan obstacle avoidance trajectory and make other behavioral decisions, and sets it as the second priority. The early warning layer's perception information is based on long-term records or sensor decision-making information currently in the warning area. This information indicates that there is a certain risk but passage is possible, and it is set as the third priority.
3. The method for decision-making on driving behavior of unmanned underground assisted transport vehicles based on multi-source information fusion according to claim 1, characterized in that, The autonomous driving mode decision based on unmanned assisted transport vehicle behavior information and safety assessment standards includes the following steps: S21. If the obstacle that suddenly appears in the perception field of the unmanned assisted transport vehicle is a sudden obstacle, then proceed to the safety assessment and use the safety assessment criteria to make driving behavior decisions; otherwise, proceed to step S22. S22, when the obstacle is not a sudden obstacle, the vehicle makes driving behavior decisions based on whether the distance between the obstacle and the front of the vehicle reaches the collision avoidance distance or obstacle avoidance distance.
4. The multi-source information fusion-based decision-making method for driving behavior of unmanned underground transport vehicles according to claim 3, characterized in that, Step S21 specifically includes the following process: The system determines whether a sudden obstacle has speed and its direction. If no speed is detected, the obstacle is considered static. It calculates the minimum braking distance and compares it with the detected obstacle's depth of field. Based on the comparison result, it autonomously switches driving modes. Specifically: When the obstacle is a static obstacle, if the braking distance is greater than or equal to the collision avoidance distance, the vehicle switches to obstacle avoidance driving mode and determines whether the obstacle is a road condition layer obstacle. If so, the vehicle makes an emergency stop until the obstacle is far away, and then automatically switches to normal driving mode or collision avoidance driving mode according to the distance. If the braking distance is less than the collision avoidance distance, the vehicle switches to collision avoidance driving mode when the braking distance is greater than or equal to the obstacle avoidance distance; and switches to normal driving mode when the braking distance is less than the obstacle avoidance distance. When an obstacle possesses both speed and acceleration, the collision time coefficient is used as the evaluation standard, and the driving mode is automatically switched based on the comparison results; specifically: Comparison of collision time coefficients and collision warning threshold Braking threshold The relative size, when When the vehicle switches to normal driving mode; when If the obstacle is not far away, the vehicle switches to collision avoidance driving mode; otherwise, it checks whether the obstacle is far away. If it is not far away, the vehicle comes to an emergency stop, reports the incident, and waits for staff to clear the obstacle. If the obstacle is far away, the vehicle switches to normal driving mode. When the destination is reached, the driving behavior decision ends.
5. The multi-source information fusion-based decision-making method for driving behavior of unmanned underground transport vehicles according to claim 3, characterized in that, Step S22 specifically includes the following process: When the onboard sensors detect that the distance between the obstacle and the unmanned assisted transport vehicle is greater than the collision avoidance distance or there is no obstacle at present, the unmanned assisted transport vehicle is in normal driving mode; When the onboard sensors detect that the distance between the obstacle and the unmanned assisted transport vehicle is less than the collision avoidance distance, the unmanned assisted transport vehicle autonomously switches to the collision avoidance driving mode. When the vehicle's onboard sensors detect an obstacle within a safe distance and before it reaches the obstacle avoidance distance, the vehicle slows down. When the onboard sensors detect an obstacle within the obstacle avoidance distance, the unmanned assisted transport vehicle autonomously switches to obstacle avoidance driving mode.
6. The multi-source information fusion-based decision-making method for driving behavior of unmanned underground transport vehicles according to claim 5, characterized in that, When the onboard sensors detect an obstacle within the obstacle avoidance distance, the unmanned assisted transport vehicle autonomously switches to obstacle avoidance driving mode, including: When the vehicle collects road condition layer perception information, and this causes the height and width of the alley to be lower than the height and width of the vehicle after the alley converges, the unmanned assisted transport vehicle will stop urgently. When the vehicle is a rail vehicle, it can only travel under the constraints of the track. At this time, the vehicle will make an emergency stop until the obstacle is far away, and then it will automatically switch to the normal driving mode or the collision avoidance driving mode according to the distance. When the vehicle is an unmanned assisted transport vehicle, the optimal obstacle avoidance path is obtained based on local path planning.
7. A multi-source information fusion-based decision-making system for the driving behavior of unmanned underground assisted transport vehicles, characterized in that: The decision-making method for driving behavior of an unmanned underground assisted transport vehicle based on multi-source information fusion as described in any one of claims 1-6, the decision system comprising a perception module, a decision module, and a control module; The perception module includes a variety of vehicle-mounted sensors, which are used to perceive and identify moving and static, near and far multimodal obstacles and road conditions to obtain multi-source perception information. The decision-making module is used to make decisions based on multi-source sensing information; The control module is used to control the underlying hardware devices to complete autonomous driving based on decision information.
Citation Information
Patent Citations
Application Methods of Virtual LiDAR Technology in Autonomous Mining Trucks
CN113759390B
Driving behavior test system and method in network connection environment based on driving simulation technology
CN116481833A
Vehicle intelligent driving behavior decision-making method and system
CN116552552A
Behavior decision-making method and system for driving of intelligent network-connected vehicle under various road conditions
CN114932918A
Decision-making system and method for underground unmanned vehicle based on multi-source information
CN115981337A
