AGV forklift anti-collision safety system

Abstract

This invention discloses an AGV forklift anti-collision safety system, belonging to the field of forklift safety control technology. The system includes a hardware detection module, a data processing module, a motion trend prediction module, an anti-collision control module, and a fault self-diagnosis module. The hardware detection module consists of front and rear 2D LiDAR and surrounding safety edges forming a dual heterogeneous redundant protection. The motion trend prediction module calculates collision time, adaptively calculates dynamic braking distance based on operating conditions, and corrects risk levels by fusing environmental semantic maps to achieve multi-dimensional collision risk prediction. The anti-collision control module generates three-level protection commands based on the prediction results, with the safety edges acting as the highest priority interruption source to unconditionally execute an emergency stop. This invention achieves 100% detection coverage without blind spots, advances the response time by 0.5 to 1 second, and adaptively adjusts the protection area according to speed and road conditions, balancing extreme safety with operational continuity, and significantly improving the safety level of unmanned operation of AGV forklifts in complex dynamic scenarios.

Description

Technical Field

[0001] This invention relates to the field of forklift safety control technology, specifically to an AGV forklift anti-collision safety system. Background Technology

[0002] With the rapid development of Industry 4.0 and intelligent warehousing and logistics, AGV forklifts, as core automated equipment for material handling and stacking, have been widely used in various warehouses and manufacturing workshops. During unmanned autonomous operation, AGV forklifts face collision risks in various complex environments, including personnel movement, equipment placement, cargo stacking, and multiple AGV intersections, placing extremely high demands on collision avoidance safety systems. Existing AGV collision avoidance solutions mainly rely on single-type sensors for obstacle detection, such as using lidar for non-contact ranging and early warning, or installing mechanical anti-collision strips for passive contact protection. Some solutions also introduce a graded early warning mechanism based on fixed distance thresholds, triggering corresponding warnings or stopping commands when obstacles enter a preset distance range. Simultaneously, some AGV systems are beginning to utilize environmental maps for path planning to avoid known static obstacle areas.

[0003] However, existing technical solutions generally have the following shortcomings:

[0004] Firstly, for dynamic obstacles, such as walking personnel or other AGVs in operation, there is a lack of ability to predict their movement intentions and future trajectories. Judging solely based on the current instantaneous distance makes it highly susceptible to collisions due to response lag.

[0005] Secondly, the protection distance threshold is usually a fixed value preset by the factory, which cannot be adaptively adjusted according to changes in the actual operating speed, load weight and ground friction coefficient of the AGV, resulting in insufficient braking distance at high speed and heavy load, or excessive protection and frequent interruption of operation at low speed and no load.

[0006] Third, the environmental map is only used for path planning and is not deeply integrated with real-time collision risk assessment. It cannot use the semantic information in the map to dynamically correct collision avoidance decisions, such as impassable areas and high-risk areas.

[0007] The aforementioned defects severely restrict the safety and operational continuity of AGV forklifts in complex and dynamic scenarios, and there is an urgent need for an intelligent anti-collision safety system with dynamic obstacle prediction, adaptive threshold adjustment and environmental semantic fusion capabilities to solve the pain points of existing technologies. Summary of the Invention

[0008] (a) Technical problems to be solved

[0009] To address the shortcomings of existing technologies, this invention provides an AGV forklift anti-collision safety system, which solves the problems mentioned in the background section.

[0010] (II) Technical Solution

[0011] To achieve the above objectives, the present invention provides the following technical solution: an AGV forklift anti-collision safety system, comprising:

[0012] The hardware detection module includes multiple lidars arranged along the AGV body and safety contact edges installed along the perimeter of the body;

[0013] The data processing module is used to receive and process obstacle information and trigger signals collected by the hardware detection module;

[0014] The motion trend prediction module is used to predict collision risks based on the processed obstacle information and AGV operating status parameters.

[0015] The motion trend prediction module is specifically configured as follows:

[0016] Acquire motion information of dynamic obstacles, and predict their predicted trajectory area within a preset time period based on the historical motion trajectory of the dynamic obstacles;

[0017] Obtain the current operating condition parameters of the AGV, including at least the operating speed, load weight and ground friction coefficient. Calculate the dynamic braking distance in real time based on the operating condition parameters, and generate multi-level dynamic protection distance thresholds based on the dynamic braking distance.

[0018] Obtain an environmental semantic map of the work scene, and correct the predicted collision risk level based on the semantic constraint information in the environmental semantic map;

[0019] The collision avoidance control module is used to generate graded protection control commands based on the corrected collision risk level and the dynamic protection distance threshold, so as to control the AGV to perform corresponding protection actions.

[0020] Furthermore, the motion trend prediction module acquires the motion information of the dynamic obstacle and predicts its predicted trajectory area within a preset time period based on the historical motion trajectory of the dynamic obstacle, including performing multi-frame continuous tracking and clustering on the point cloud data collected by the lidar, identifying the dynamic obstacle and calculating its motion speed and direction.

[0021] Based on the relative distance between the AGV and the dynamic obstacle and relative velocity According to the formula Real-time calculation of collision time When the collision time is less than a preset threshold, it is determined that there is an emergency collision risk.

[0022] Furthermore, the motion trend prediction module acquires the current operating parameters of the AGV and calculates the dynamic braking distance in real time. Based on the dynamic braking distance, it generates multi-level dynamic protection distance thresholds, including acquiring the current operating speed of the AGV through the AGV motion control system or on-board sensors. And obtain the ground friction coefficient of the current working area through the communication interface. The dynamic emergency braking distance is calculated in real time according to the following formula. :

[0023] ;

[0024] in It is the acceleration due to gravity. To set a safety margin, the dynamic emergency braking distance is used as the dynamic emergency stopping distance threshold, and this threshold is multiplied by a preset gain coefficient to serve as the dynamic deceleration distance threshold.

[0025] Furthermore, the motion trend prediction module acquires an environmental semantic map of the work scene and corrects the predicted collision risk level based on semantic constraint information, including:

[0026] The predicted motion trajectory of the AGV is compared with the environmental semantic map;

[0027] When the predicted trajectory points to an impassable area, the collision risk level is corrected to the highest level.

[0028] When the predicted motion trajectory enters or approaches a semantic region marked as high risk, the weight of the collision risk level is increased.

[0029] Furthermore, the lidar in the hardware detection module is a 2D lidar installed at the front and rear of the AGV respectively, used to achieve 360° non-contact obstacle detection; the safety contact edge is a flexible contact edge with an embedded trigger sensor, installed around the AGV; when the data processing module determines that the lidar function is malfunctioning or there is a detection blind spot, it activates the highest priority response mechanism of the safety contact edge, so that it directly outputs an emergency stop command when triggered.

[0030] Furthermore, the collision avoidance control module executes graded protection control commands, including three levels:

[0031] Level 1 Protection Command: When an obstacle exceeds the dynamic deceleration distance threshold and there is no risk of emergency collision, trigger an audible and visual warning and maintain the current speed;

[0032] Level 2 protection command: When an obstacle enters between the dynamic deceleration distance threshold and the dynamic emergency stop distance threshold, or when there is a potential collision risk, an audible and visual warning is triggered and the AGV is controlled to decelerate to less than 50% of its current speed;

[0033] Level 3 protection command: When an obstacle enters within the dynamic emergency stop distance threshold, or there is an emergency collision risk, or the safety edge is triggered, an emergency stop shall be executed immediately.

[0034] Furthermore, the anti-collision control module is configured such that when the safety edge is triggered, the level 3 protection command is executed first, regardless of the current collision risk level.

[0035] Furthermore, it also includes a fault self-checking module, which is used to periodically check the status of each module. When any abnormality is detected in any module, an alarm is triggered and the AGV is controlled to stop, ensuring system failure-oriented safety.

[0036] (III) Beneficial Effects

[0037] 1. 360° non-contact predictive detection is achieved through front and rear 2D LiDAR, combined with safety contact edges continuously installed along the vehicle's perimeter to form a contact-based bottom-line protection system. The two are deeply coupled in control logic. When the LiDAR fails or a detection blind spot exists, the safety contact edge automatically activates the highest priority response mechanism, completely eliminating the blind spots and failure risks inherent in single-sensor solutions, and achieving 100% physical coverage of obstacle detection.

[0038] 2. By continuously tracking multi-frame point clouds and using Kalman filtering to identify dynamic obstacles, and by introducing quantitative calculation of collision time, the urgency of collision is improved from instantaneous distance judgment to time-dimensional prediction based on relative speed and distance. This advances the response time by 0.5 to 1 second, providing the AGV with sufficient braking and avoidance time, effectively solving the pain point of delayed response when facing moving targets in traditional solutions.

[0039] 3. By changing the protection distance threshold, the protection distance threshold is no longer a fixed value preset by the factory. Instead, it is dynamically calculated in real time based on the AGV's current operating speed, load weight, and ground friction coefficient, according to kinematic formulas. The protection area automatically expands when the AGV is running at high speed and under heavy load or on a low-friction surface, and shrinks accordingly when the AGV is running at low speed and under no load. This avoids insufficient braking under extreme working conditions and reduces frequent work interruptions caused by over-protection, achieving a dynamic balance between safety and work efficiency.

[0040] 4. By deeply integrating real-time collision risk assessment with high-precision semantic maps, the system proactively corrects the collision risk level when the predicted trajectory points to impassable areas such as shelves or walls, or enters high-risk semantic areas. This elevates the system from simple geometric perception to an environmental understanding of the operational scenario structure, significantly enhancing adaptability and decision-making reliability in complex warehouse layouts.

[0041] 5. Through a three-level protection logic, warning, deceleration, and emergency stop commands are dynamically output according to the risk level. The safety contact edge is the highest priority interruption source. Regardless of the current decision status, an emergency stop is unconditionally executed when triggered. Under the premise of ensuring extreme safety, the continuity and efficiency of AGV operation are maintained to the maximum extent. Attached Figure Description

[0042] Figure 1 This is a system structure block diagram of the present invention;

[0043] Figure 2 This is a flowchart of the collision time prediction process in the motion trend prediction module of this invention;

[0044] Figure 3 This is a schematic diagram of the motion trend prediction scenario of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Please see Figures 1 to 3 As shown, the embodiments of the present invention provide the following technical solutions:

[0047] This embodiment discloses an AGV forklift anti-collision safety system, which specifically includes the following modules:

[0048] 1. Hardware detection module, including multiple lidars arranged along the AGV body and safety contact edges installed along the perimeter of the body, adopts a dual heterogeneous redundancy protection design with non-contact early warning and contact-based backup.

[0049] Specifically, in this embodiment, two 2D lidar units are selected and installed at the front and rear center of the AGV forklift, respectively. This front-rear layout achieves 360° omnidirectional non-contact obstacle detection coverage in the horizontal direction, excluding areas obstructed by the vehicle itself. The lidar emits laser pulses and receives reflected signals, accurately calculating the distance, orientation, and contour information of obstacles based on the triangulation principle, and outputs point cloud data. Its main technical parameters are: detection distance 0.1-10 meters, ranging accuracy ≤ ±5mm, and scanning frequency ≥ 10Hz. To ensure stable operation in complex industrial environments, the lidar adopts a hardware design resistant to dust and strong light interference, and, in conjunction with the filtering algorithm of the subsequent data processing module, can effectively filter out invalid signals caused by suspended particles, reflected light spots, etc.

[0050] The safety contact edge is made of flexible rubber and has a trigger-type sensor embedded along its entire length. It is continuously installed along the front, rear, and side edges of the AGV forklift, forming a complete circumferential contact protection ring. The safety contact edge is triggered when subjected to an external pressure of 0.5-1N, with a response time of ≤10ms, sending a switch trigger signal to the data processing module. This safety contact edge has excellent wear resistance, deformation resistance, and waterproof and dustproof properties, and has a self-resetting function after the pressure is removed. As the last physical safety line of the system, the safety contact edge provides backup protection in case of lidar failure, detection blind spots, or algorithm failure to respond in time.

[0051] 2. Data Processing Module: The data processing module is integrated into the AGV's onboard controller. The onboard controller is based on an ARM architecture with a main frequency of no less than 1.5GHz and memory of no less than 2GB. It is used to receive and process obstacle information and trigger signals collected by the hardware detection module, and perform preprocessing operations. For LiDAR point cloud data, Gaussian filtering algorithm is used for noise reduction to remove outliers, and effective point cloud is clustered and analyzed to extract the contour, distance, and motion information of obstacles. For safety edge signals, anti-shake processing is performed to eliminate momentary false triggers. When the data processing module determines that the LiDAR function is malfunctioning or has a detection blind zone based on abnormal point cloud data density, continuous absence of effective echo, or LiDAR self-diagnosis interface, the highest priority response mechanism of the safety edge is activated, so that it directly outputs an emergency stop command when triggered.

[0052] 3. Motion Trend Prediction Module. The motion trend prediction module is the core intelligent decision-making unit of this system, used to predict collision risks based on processed obstacle information and AGV operating status parameters.

[0053] The motion trend prediction module is specifically configured to have the following three sub-functions:

[0054] (1) Obtain the motion information of the dynamic obstacle and predict its predicted trajectory area within a preset time period based on the historical motion trajectory of the dynamic obstacle;

[0055] Specifically, this addresses dynamic collision risk scenarios such as personnel movement and the movement of other AGVs. The motion trend prediction module acquires the motion information of dynamic obstacles and predicts their predicted trajectory area within a preset time based on their historical motion trajectories. This includes performing multi-frame continuous tracking and clustering of point cloud data collected by the lidar, identifying dynamic obstacles through inter-frame target matching, and estimating their motion speed and direction using a Kalman filter algorithm.

[0056] Based on this, the motion trend prediction module obtains the relative distance between the AGV and the dynamic obstacle. and relative velocity According to the formula Real-time calculation of collision time When the collision time is less than a preset threshold, an emergency collision risk is determined, and the warning level is directly upgraded to give the system additional response lead time. Specifically, this embodiment presets two levels. Thresholds: First threshold T1 and second threshold T2, where T1 > T2. When At that time, it was determined that there was no risk of an emergency collision; when When, it is determined that the area has entered a potential collision risk zone; when At that time, it was determined that there was an emergency collision risk.

[0057] (2) Obtain the current operating condition parameters of the AGV, including at least the operating speed, load weight and ground friction coefficient, calculate the dynamic braking distance in real time based on the operating condition parameters, and generate multi-level dynamic protection distance thresholds based on the dynamic braking distance;

[0058] Specifically, to address the issue that fixed protection distances cannot adapt to different operating conditions, the motion trend prediction module acquires the current operating parameters of the AGV and calculates the dynamic braking distance in real time. Based on the dynamic braking distance, it generates multi-level dynamic protection distance thresholds, including acquiring the current operating speed of the AGV through the AGV motion control system or onboard sensors. And obtain the ground friction coefficient of the current working area through the communication interface. The dynamic emergency braking distance is calculated in real time according to the following formula. :

[0059] ;

[0060] in, It is the acceleration due to gravity. To preset a safety margin; the dynamic emergency braking distance As a dynamic emergency stopping distance threshold, and will After being multiplied by a preset gain coefficient, it serves as the dynamic deceleration distance threshold. This mechanism ensures that the protected area can adaptively expand under extreme conditions such as high speed, heavy load, and low-friction road surfaces, always reserving sufficient braking distance for the AGV.

[0061] (3) Obtain the environmental semantic map of the work scene, and correct the predicted collision risk level based on the semantic constraint information in the environmental semantic map;

[0062] Specifically, the motion trend prediction module acquires an environmental semantic map of the work scene and corrects the predicted collision risk level based on semantic constraint information to improve the ability to predict static risk areas, including:

[0063] The semantic map is pre-labeled with semantic tags such as passable areas, impassable areas (shelves, walls), and mixed pedestrian areas. The AGV's current predicted movement trajectory is compared with the environmental semantic map in real time.

[0064] When the predicted trajectory points to an impassable area, the collision risk level is corrected to the highest level, regardless of whether the sensor detects an obstacle.

[0065] When the predicted motion trajectory enters or approaches a semantic region marked as high risk, such as a human-machine mixed area, the weight coefficient of the collision risk level corresponding to that region is actively increased, so that the system moves in that region with a more cautious strategy.

[0066] 4. Collision Avoidance Control Module. The collision avoidance control module, acting as a decision-making and execution unit, generates graded protection control commands based on the revised collision risk level and the dynamic protection distance threshold. It seamlessly integrates with the AGV motion control system via a communication linkage module. This module uses the Profinet protocol to communicate with the AGV motion control system and the onboard display terminal, with a communication latency of ≤10ms, ensuring real-time command transmission and timely control of the AGV to execute corresponding protective actions.

[0067] Specifically, the collision avoidance control module executes three levels of graded protection control commands:

[0068] Level 1 Protection Command: When the obstacle exceeds the dynamic deceleration distance threshold and the collision time... When no emergency collision risk is determined, the system triggers an industrial-grade audible and visual alarm (volume ≥80dB) to issue a warning, and displays the obstacle's location and distance information on a 7-inch vehicle-mounted touch terminal. The AGV maintains its current operating speed and continues to operate, ensuring the continuity of operations.

[0069] Level 2 protection command: When an obstacle enters between the dynamic deceleration distance threshold and the dynamic emergency stopping distance threshold, or the collision time... Upon entering a potentially risky area, trigger an audible and visual warning and control the AGV to slow down to below 50% of its current speed. If necessary, perform minor steering adjustments based on the obstacle's location to actively avoid it. (See attached diagram.) Figure 3 .

[0070] Level 3 protection command: When an obstacle enters within the dynamic emergency braking distance threshold, or the collision time... When an emergency collision risk is detected or a safety edge is triggered, an emergency stop is immediately executed, power output is cut off and the brakes are activated to ensure braking is completed within 0.3 meters. At the same time, the audible and visual alarms continue until the obstacle is cleared and the system is manually reset.

[0071] Furthermore, the safety edge trigger is set as the highest priority interrupt source in the system. Regardless of the current collision risk level and distance assessment results, once the safety edge signal is triggered, the collision avoidance control module will directly bypass all other decision logic and unconditionally execute the Level 3 protection command, ensuring that the system is guided to a safe state the moment physical contact occurs.

[0072] 5. Fault Self-Check Module. The fault self-check module is built into the system and periodically checks the status of each critical module at preset time intervals. The checks include the LiDAR communication status, the continuity of the safety contact circuit, the operating status of the data processing module, and the quality of the communication link. When any abnormality is detected in any module, the self-check module immediately triggers an alarm and forces the AGV to stop via the anti-collision control module. This achieves the functional safety design requirements for system failure-oriented safety and avoids protection gaps caused by malfunctions in the anti-collision system itself.

[0073] Overall workflow description:

[0074] With the coordinated operation of the above modules, this system forms a complete closed-loop intelligent collision avoidance workflow, realizing full automation from environmental perception, risk prediction, decision control to safe execution.

[0075] System power-on and self-test phase

[0076] After the AGV forklift is powered on and started, the anti-collision safety system first enters the self-test program. The fault self-test module, according to a preset inspection cycle, sequentially performs a comprehensive test on the communication link and data transmission status of the LiDAR, the circuit continuity and signal response of the safety contact edge, the computing resources and memory status of the data processing module, and the bus connection quality of the communication linkage module. If all modules are normal, the system enters standby ready mode and sends a feedback signal to the AGV motion control system indicating that the anti-collision system is ready, allowing the AGV to execute its work tasks. If any abnormality is detected in any module, such as no data transmission from the LiDAR or an open circuit in the safety contact edge, the fault self-test module immediately triggers an alarm to issue a fault alarm. Simultaneously, it sends a forced stop command to the motion control system through the anti-collision control module, prohibiting the AGV from starting and running. This achieves the functional safety design goal of failure-oriented safety, eliminating the risk of safety protection deficiencies caused by faults in the anti-collision system itself.

[0077] Normal working cycle phase

[0078] In the perception phase, front and rear 2D LiDAR continuously scans a range of 0.1-10 meters around the vehicle at a frequency of no less than 10Hz, generating point cloud data in real time; the safety edge is always in standby monitoring mode. After receiving the point cloud data, the data processing module uses a Gaussian filtering algorithm to reduce noise and filter out interference such as dust and strong light, and performs clustering and segmentation on the filtered point cloud to extract the geometric dimensions and relative positions of obstacles; at the same time, the safety edge signal is subjected to dual hardware and software anti-shake processing to eliminate instantaneous false triggers caused by vibration.

[0079] In the core prediction stage, the motion trend prediction module performs three tasks in parallel: First, it performs target tracking and inter-frame matching on multiple consecutive frames of point cloud data, identifies dynamic obstacles, and estimates their speed and direction using Kalman filtering, according to the formula... Collision time is calculated in real time, and based on two preset threshold levels, it determines whether there is no emergency collision risk, a potential collision risk, or an emergency collision risk, advancing the response time by 0.5 to 1 second, effectively solving the problem of delayed response to moving targets. Secondly, the current operating speed of the AGV is acquired in real time. coefficient of friction with the ground According to the formula The system calculates dynamic emergency braking distance and generates dynamic deceleration distance thresholds, enabling the protected area to adaptively adjust according to operating conditions, balancing safety during high-speed heavy loads with operational continuity during low-speed no-load operations. Thirdly, the system compares the AGV's predicted trajectory with a semantic map, directly correcting it to the highest risk level when pointing to impassable areas, and increasing the risk weight coefficient when approaching high-risk areas, thus endowing the system with environmental understanding of the operational scenario structure and robust decision-making capabilities.

[0080] In the decision-making and execution phase, the collision avoidance control module, based on the above evaluation results, issues tiered protection commands to the AGV motion control system via the Profinet bus with a delay of less than 10 milliseconds: Level 1 protection triggers an audible and visual warning and maintains the current speed; Level 2 protection triggers an audible and visual alarm and reduces speed to below 50% of the current speed, with minor steering adjustments for avoidance if necessary; Level 3 protection immediately cuts off power and applies emergency braking to ensure the vehicle stops within 0.3 meters. The safety edge, as the highest priority interruption source, unconditionally executes Level 3 protection upon triggering; when the lidar fails or a blind spot exists, the safety edge automatically activates the highest priority response, achieving heterogeneous redundancy and complementarity among sensors.

[0081] Fault handling and system recovery phase

[0082] Throughout the AGV operation, the fault self-check module continuously performs background inspections of each key module at preset intervals. When the inspection detects any communication interruption, data anomaly, or hardware failure in any module, the system immediately triggers a fault alarm and issues an emergency stop command to the motion control system via the anti-collision control module, forcing the AGV to stop running and preventing continued operation while the anti-collision function is missing. The system can only re-enter the normal working cycle after the fault has been eliminated and manually confirmed for reset.

[0083] In summary, this system constructs a blind-spot-free, dual heterogeneous redundancy protection system by combining 360° non-contact prediction with lidar and comprehensive safety contact edge coverage, achieving 100% physical coverage of obstacle detection. By introducing a collaborative intelligent prediction mechanism involving TTC collision time calculation, dynamic braking distance adaptive adjustment, and semantic map risk correction, it achieves a technological leap from passive response to proactive prediction. Through a three-level hierarchical protection logic and the design of the highest priority instruction at the safety contact edge, it ensures extreme safety while maintaining operational continuity and efficiency. The modular architecture and fault self-checking failure safety mechanism ensure the system's reliability, maintainability, and ease of deployment, significantly improving the safety level of AGV forklifts in complex and dynamic scenarios such as intelligent warehouses and factory workshops.

[0084] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0085] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0086] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An AGV forklift anti-collision safety system, characterized in that, include: The hardware detection module includes multiple lidars arranged along the AGV body and safety contact edges installed along the perimeter of the body; The data processing module is used to receive and process obstacle information and trigger signals collected by the hardware detection module; The motion trend prediction module is used to predict collision risks based on the processed obstacle information and AGV operating status parameters. The motion trend prediction module is specifically configured as follows: Acquire motion information of dynamic obstacles, and predict their predicted trajectory area within a preset time period based on the historical motion trajectory of the dynamic obstacles; Obtain the current operating condition parameters of the AGV, including at least the operating speed, load weight and ground friction coefficient. Calculate the dynamic braking distance in real time based on the operating condition parameters, and generate multi-level dynamic protection distance thresholds based on the dynamic braking distance. Obtain an environmental semantic map of the work scene, and correct the predicted collision risk level based on the semantic constraint information in the environmental semantic map; The collision avoidance control module is used to generate graded protection control commands based on the corrected collision risk level and the dynamic protection distance threshold, so as to control the AGV to perform corresponding protection actions.

2. The AGV forklift anti-collision safety system of claim 1, wherein, The motion trend prediction module acquires the motion information of dynamic obstacles and predicts their predicted trajectory area within a preset time based on the historical motion trajectory of the dynamic obstacles. This includes performing multi-frame continuous tracking and clustering on the point cloud data collected by the lidar, identifying the dynamic obstacles, and calculating their motion speed and direction. According to the relative distance between the AGV and the dynamic obstacle and the relative speed , the collision time is calculated in real time according to the formula When the collision time is less than a preset threshold, it is determined that there is an emergency collision risk.​ 3. The AGV forklift anti-collision safety system according to claim 1, characterized in that, The motion trend prediction module acquires the current operating parameters of the AGV and calculates the dynamic braking distance in real time. Based on the dynamic braking distance, it generates multi-level dynamic protection distance thresholds, including acquiring the current operating speed of the AGV through the AGV motion control system or on-board sensors. And obtain the ground friction coefficient of the current working area through the communication interface. The dynamic emergency braking distance is calculated in real time according to the following formula. : ； in It is the acceleration due to gravity. To set a safety margin, the dynamic emergency braking distance is used as the dynamic emergency stopping distance threshold, and this threshold is multiplied by a preset gain coefficient to serve as the dynamic deceleration distance threshold.

4. The AGV forklift anti-collision safety system according to claim 1, characterized in that, The motion trend prediction module acquires an environmental semantic map of the work scene and corrects the predicted collision risk level based on semantic constraint information, including: The predicted motion trajectory of the AGV is compared with the environmental semantic map; When the predicted trajectory points to an impassable area, the collision risk level is corrected to the highest level. When the predicted motion trajectory enters or approaches a semantic region marked as high risk, the weight of the collision risk level is increased.

5. The AGV forklift anti-collision safety system according to any one of claims 1 to 4, characterized in that, The LiDAR in the hardware detection module is a 2D LiDAR installed at the front and rear of the AGV respectively, used to achieve 360° non-contact obstacle detection; the safety contact edge is a flexible contact edge with an embedded trigger sensor installed around the AGV; when the data processing module determines that the LiDAR is malfunctioning or has a detection blind spot, it activates the highest priority response mechanism of the safety contact edge, so that it directly outputs an emergency stop command when triggered.

6. The AGV forklift anti-collision safety system according to claim 1, characterized in that, The collision avoidance control module executes three levels of graded protection control commands: Level 1 Protection Command: When an obstacle exceeds the dynamic deceleration distance threshold and there is no risk of emergency collision, trigger an audible and visual warning and maintain the current speed; Level 2 protection command: When an obstacle enters between the dynamic deceleration distance threshold and the dynamic emergency stop distance threshold, or when there is a potential collision risk, an audible and visual warning is triggered and the AGV is controlled to decelerate to less than 50% of its current speed; Level 3 protection command: When an obstacle enters within the dynamic emergency stop distance threshold, or there is an emergency collision risk, or the safety edge is triggered, an emergency stop shall be executed immediately.

7. The AGV forklift anti-collision safety system according to claim 6, characterized in that, The collision avoidance control module is configured such that when the safety edge is triggered, the level 3 protection command is executed first, regardless of the current collision risk level.

8. The AGV forklift anti-collision safety system according to claim 1, characterized in that, It also includes a fault self-checking module, which is used to periodically check the status of each module. When any abnormality is detected in any module, an alarm is triggered and the AGV is controlled to stop, ensuring system failure-oriented safety.

Images