Industrial robot anti-collision early warning method and system

Abstract

The invention provides an industrial robot anti-collision early warning method and system. The method comprises the following steps: environment perception and multi-source data fusion; predicting a dynamic obstacle track; robot posture real-time modeling is carried out; carrying out collision risk joint assessment; dynamically dividing early warning levels; intelligent planning of an obstacle avoidance path; dynamically compensating joint torque; multi-stage early warning collaborative response is carried out; performing iterative evaluation on the early warning effect; and self-evolving the anti-collision knowledge base. The three key technical problems in the industrial robot anti-collision field are solved, dependence on high-cost physical sensors (such as a torque sensor and a laser radar) is eliminated, collision detection of any position of the whole robot is achieved, and the cost is reduced by 60% or above. The limitation of a static obstacle model is overcome, and real-time track prediction and risk grading of dynamic targets such as moving personnel and conveying belt materials are achieved. Through multi-source data fusion and a risk probability model, the false alarm rate is reduced to 5% or below, and meanwhile it is guaranteed that the missing report rate is 0.

Description

Technical Field

[0001] This invention specifically relates to a collision avoidance warning method and system for industrial robots. Background Technology

[0002] Industrial robots are widely used in automated production, but they are prone to collisions in complex and dynamic environments. Existing technologies mainly rely on the following solutions and have the following limitations:

[0003] Force / torque sensors (CN202310897606.9) or wrist force sensors detect collision forces, but their detection range is limited to the installation location and cannot cover the whole body, and they are expensive.

[0004] Radar distance sensors (CN202311194946.1) need to be exposed to the external environment and are susceptible to dust and electromagnetic interference, which leads to an increased false alarm rate.

[0005] The enclosing sphere model (CN116197912A) simplifies obstacles into spheres, making it difficult to accurately fit the contours of irregularly shaped obstacles, and it does not consider the dynamic trajectory of obstacles.

[0006] Joint torque monitoring (CN202310897606.9) alarms by providing feedback on torque deviation, but it is susceptible to interference from mechanical noise and sudden acceleration and deceleration, with a false trigger rate of over 15%.

[0007] Traditional solutions require 100-200ms from the occurrence of a collision to braking, which cannot cope with high-speed moving robots (such as welding robots with a speed ≥2m / s).

[0008] In summary, this application proposes an anti-collision warning method and system for industrial robots to solve the above problems. Summary of the Invention

[0009] The purpose of this invention is to address the shortcomings of existing technologies by providing an industrial robot collision avoidance and early warning method that can effectively solve the aforementioned problems.

[0010] To achieve the above requirements, the technical solution adopted by the present invention is: to provide an industrial robot collision avoidance warning method, which includes the following steps:

[0011] S1: Steps for environmental perception and multi-source data fusion;

[0012] S2: Steps for predicting dynamic obstacle trajectories;

[0013] S3: Steps for real-time robot pose modeling;

[0014] S4: Steps for conducting a joint collision risk assessment;

[0015] S5: Steps for dynamically classifying early warning levels;

[0016] S6: Steps for intelligent obstacle avoidance path planning;

[0017] S7: Steps for performing dynamic compensation of joint torque;

[0018] S8: Steps for multi-level early warning and coordinated response;

[0019] S9: Steps for iterative evaluation of early warning effectiveness;

[0020] S10: Steps for self-evolving the collision avoidance knowledge base.

[0021] Preferably, step S1 specifically includes:

[0022] The robot collects environmental point cloud data using an RGB-D camera mounted on its body, while simultaneously reading joint angle and angular velocity signals from the motor encoders and real-time current values ​​from the servo system. The point cloud data is used to construct a 3D grid map of the environment, and the joint data is used to analyze the robot's posture. A Kalman filter is then used to fuse the visual and joint data to eliminate noise from a single sensor (such as point cloud distortion caused by changes in lighting).

[0023] Preferably, step S2 specifically includes:

[0024] Based on the grid map output in step 1, the YOLOv7 model is used to identify moving targets (people, AGVs, etc.), and the LSTM network is used to predict their movement trajectory within the next 3 seconds. The LSTM network learns historical trajectory sequences (sampling frequency 30Hz) and outputs a predicted path in the form of a probability distribution with a confidence level ≥90%.

[0025] Preferably, step S3 specifically includes:

[0026] Based on the robot's URDF model and the joint data from step 1, a digital twin is constructed, and the real-time bounding box (OBB-Oriented Bounding Box) of each link is calculated. The OBB model dynamically adjusts its orientation based on principal component analysis (PCA), achieving a 40% improvement in accuracy compared to a fixed bounding sphere (CN116197912A).

[0027] Preferably, step S4 specifically includes:

[0028] The obstacle prediction trajectory from step 2 and the robot OBB model from step 3 are input into the GJK algorithm to calculate the minimum Euclidean distance and generate the collision probability by combining the relative velocity:

[0029] ;

[0030] Where d is the real-time distance, v is the relative velocity, and t_{resp} is the system response time; the sigmoid function is used to map the physical parameters to probability values, and an early warning is triggered when P_{collision} > 0.7.

[0031] Preferably, step S5 specifically includes:

[0032] Three levels of warning are defined based on collision probability and damage severity:

[0033] Level 1: 0.7 ≤ P < 0.85, audible and visual warning;

[0034] Level 2: 0.85 ≤ P < 0.95, reduce speed to 50%;

[0035] Level 3: P ≥ 0.95, emergency braking;

[0036] A tiered strategy avoids frequent downtime and ensures production efficiency.

[0037] Preferably, step S6 specifically includes:

[0038] In Level 2 warning mode, the RRT* algorithm is used to generate a collision-free path, prioritizing the optimization of the trajectory with the smallest change in joint angle. RRT* optimizes the path cost based on random sampling to ensure smooth trajectory and minimum energy consumption.

[0039] Preferably, step S7 specifically includes:

[0040] When executing the obstacle avoidance path, the feedback torque of each joint is collected, and a compensation current is generated by the PID controller to suppress overshoot; the compensation current ΔI = K_p · e + K_i \int e dt, where e is the deviation between the theoretical torque and the feedback torque, to prevent secondary collisions caused by inertia during obstacle avoidance.

[0041] Preferably, step S8 specifically includes:

[0042] Activate the corresponding device based on the warning level:

[0043] Level 1: Activate the robot's warning lights;

[0044] Level 2: Send deceleration logs to the MES system;

[0045] Level 3: Triggers the safety relay to cut off the drive power;

[0046] Hardware response latency is controlled within 20ms, meeting the SIL 2 safety level.

[0047] Step S9 specifically includes: recording the collision probability, actual distance deviation, and response time for each warning; updating the weight parameters of the LSTM network; and optimizing the prediction model based on the PPO algorithm of reinforcement learning to reduce model drift caused by environmental changes.

[0048] Step S10 specifically includes: storing the evaluation results from step 9 into the graph database Neo4j, establishing "scene-action-effect" association rules; when a similar scene reappears, directly calling the historical best strategy, reducing the planning time to 100ms.

[0049] An industrial robot collision avoidance warning system is provided, which uses any of the industrial robot collision avoidance warning methods described above.

[0050] The advantages of this industrial robot collision avoidance and early warning method are as follows:

[0051] By setting Attached Figure Description

[0052] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, use the same reference numerals to denote the same or similar parts. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0053] Figure 1 A schematic flowchart of an industrial robot collision avoidance warning method according to an embodiment of this application is shown. Detailed Implementation

[0054] 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 specific embodiments.

[0055] In the following description, references to "an embodiment," "an embodiment," "an example," "example," etc., indicate that the described embodiment or example may include a particular feature, structure, characteristic, property, element, or limitation, but not every embodiment or example necessarily includes that particular feature, structure, characteristic, property, element, or limitation. Furthermore, the repeated use of the phrase "an embodiment according to this application," while possibly referring to the same embodiment, does not necessarily refer to the same embodiment.

[0056] For simplicity, certain technical features known to those skilled in the art are omitted in the following description.

[0057] According to one embodiment of this application, an anti-collision warning method for industrial robots is provided, such as... Figure 1 As shown, it includes the following steps:

[0058] S1: Steps for environmental perception and multi-source data fusion;

[0059] S2: Steps for predicting dynamic obstacle trajectories;

[0060] S3: Steps for real-time robot pose modeling;

[0061] S4: Steps for conducting a joint collision risk assessment;

[0062] S5: Steps for dynamically classifying early warning levels;

[0063] S6: Steps for intelligent obstacle avoidance path planning;

[0064] S7: Steps for performing dynamic compensation of joint torque;

[0065] S8: Steps for multi-level early warning and coordinated response;

[0066] S9: Steps for iterative evaluation of early warning effectiveness;

[0067] S10: Steps for self-evolving the collision avoidance knowledge base.

[0068] According to one embodiment of this application, step S1 of the industrial robot collision avoidance warning method specifically includes:

[0069] The robot collects environmental point cloud data using an RGB-D camera mounted on its body, while simultaneously reading joint angle and angular velocity signals from the motor encoders and real-time current values ​​from the servo system. The point cloud data is used to construct a 3D grid map of the environment, and the joint data is used to analyze the robot's posture. A Kalman filter is then used to fuse the visual and joint data to eliminate noise from a single sensor (such as point cloud distortion caused by changes in lighting).

[0070] According to one embodiment of this application, step S2 of the industrial robot collision avoidance warning method specifically includes:

[0071] Based on the grid map output in step 1, the YOLOv7 model is used to identify moving targets (people, AGVs, etc.), and the LSTM network is used to predict their movement trajectory within the next 3 seconds. The LSTM network learns historical trajectory sequences (sampling frequency 30Hz) and outputs a predicted path in the form of a probability distribution with a confidence level ≥90%.

[0072] According to one embodiment of this application, step S3 of the industrial robot collision avoidance warning method specifically includes:

[0073] Based on the robot's URDF model and the joint data from step 1, a digital twin is constructed, and the real-time bounding box (OBB-Oriented Bounding Box) of each link is calculated. The OBB model dynamically adjusts its orientation based on principal component analysis (PCA), achieving a 40% improvement in accuracy compared to a fixed bounding sphere (CN116197912A).

[0074] According to one embodiment of this application, step S4 of the industrial robot collision avoidance warning method specifically includes:

[0075] The obstacle prediction trajectory from step 2 and the robot OBB model from step 3 are input into the GJK algorithm to calculate the minimum Euclidean distance and generate the collision probability by combining the relative velocity:

[0076] ;

[0077] Where d is the real-time distance, v is the relative velocity, and t_{resp} is the system response time; the sigmoid function is used to map the physical parameters to probability values, and an early warning is triggered when P_{collision} > 0.7.

[0078] According to one embodiment of this application, step S5 of the industrial robot collision avoidance warning method specifically includes:

[0079] Three levels of warning are defined based on collision probability and damage severity:

[0080] Level 1: 0.7 ≤ P < 0.85, audible and visual warning;

[0081] Level 2: 0.85 ≤ P < 0.95, reduce speed to 50%;

[0082] Level 3: P ≥ 0.95, emergency braking;

[0083] A tiered strategy avoids frequent downtime and ensures production efficiency.

[0084] According to one embodiment of this application, step S6 of the industrial robot collision avoidance warning method specifically includes:

[0085] In Level 2 warning mode, the RRT* algorithm is used to generate a collision-free path, prioritizing the optimization of the trajectory with the smallest change in joint angle. RRT* optimizes the path cost based on random sampling to ensure smooth trajectory and minimum energy consumption.

[0086] According to one embodiment of this application, step S7 of the industrial robot collision avoidance warning method specifically includes:

[0087] When executing the obstacle avoidance path, the feedback torque of each joint is collected, and a compensation current is generated by the PID controller to suppress overshoot; the compensation current ΔI = K_p · e + K_i \int e dt, where e is the deviation between the theoretical torque and the feedback torque, to prevent secondary collisions caused by inertia during obstacle avoidance.

[0088] According to one embodiment of this application, step S8 of the industrial robot collision avoidance warning method specifically includes:

[0089] Activate the corresponding device based on the warning level:

[0090] Level 1: Activate the robot's warning lights;

[0091] Level 2: Send deceleration logs to the MES system;

[0092] Level 3: Triggers the safety relay to cut off the drive power;

[0093] Hardware response latency is controlled within 20ms, meeting the SIL 2 safety level.

[0094] According to one embodiment of this application, step S9 of the industrial robot collision avoidance warning method specifically includes: recording the collision probability, actual distance deviation, and response time for each warning; updating the weight parameters of the LSTM network; and optimizing the prediction model based on the PPO algorithm of reinforcement learning to reduce model drift caused by environmental changes.

[0095] According to one embodiment of this application, step S10 of the industrial robot collision avoidance warning method specifically includes:

[0096] The evaluation results from step 9 are stored in the graph database Neo4j, and "scene-action-effect" association rules are established. When a similar scene reappears, the historical best strategy is directly invoked, reducing the planning time to 100ms.

[0097] According to one embodiment of this application, an industrial robot collision avoidance warning system is provided, using any of the above-described industrial robot collision avoidance warning methods.

[0098] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the claims.

Claims

1. A collision avoidance warning method for industrial robots, characterized in that, Includes the following steps: S1: Steps for environmental perception and multi-source data fusion; S2: Steps for predicting dynamic obstacle trajectories; S3: Steps for real-time robot pose modeling; S4: Steps for conducting a joint collision risk assessment; S5: Steps for dynamically classifying early warning levels; S6: Steps for intelligent obstacle avoidance path planning; S7: Steps for performing dynamic compensation of joint torque; S8: Steps for multi-level early warning and coordinated response; S9: Steps for iterative evaluation of early warning effectiveness; S10: Steps for self-evolution of the collision avoidance knowledge base.

2. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S1 specifically includes: The robot collects environmental point cloud data using an RGB-D camera mounted on its body, and simultaneously reads the joint angle and angular velocity signals of the motor encoder and the real-time current value of the servo system. The point cloud data is used to construct a 3D grid map of the environment, and the joint data is used to analyze the robot's posture. The visual and joint data are fused using a Kalman filter to eliminate noise from a single sensor.

3. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S2 specifically includes: Based on the grid map output in step 1, the YOLOv7 model is used to identify moving targets, and the LSTM network is used to predict their trajectory within the next 3 seconds. The LSTM network learns the historical trajectory sequence and outputs the predicted path in the form of a probability distribution with a confidence level of ≥90%.

4. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S3 specifically includes: Based on the robot's URDF model and the joint data from step 1, a digital twin is constructed, and the real-time bounding box of each link is calculated. The OBB model dynamically adjusts its direction based on principal component analysis, achieving a 40% improvement in accuracy compared to a fixed bounding sphere.

5. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S4 specifically includes: The obstacle prediction trajectory from step 2 and the robot OBB model from step 3 are input into the GJK algorithm to calculate the minimum Euclidean distance and generate the collision probability by combining the relative velocity: ； Where d is the real-time distance, v is the relative velocity, and t_{resp} is the system response time; the sigmoid function is used to map the physical parameters to probability values, and an early warning is triggered when P_{collision} > 0.

7.

6. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S5 specifically includes: Three levels of warning are defined based on collision probability and damage severity: Level 1: 0.7 ≤ P < 0.85, audible and visual warning; Level 2: 0.85 ≤ P < 0.95, reduce speed to 50%; Level 3: P ≥ 0.95, emergency braking; A tiered strategy avoids frequent downtime and ensures production efficiency.

7. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S6 specifically includes: In Level 2 warning mode, the RRT* algorithm is used to generate a collision-free path, prioritizing the optimization of the trajectory with the smallest change in joint angle. RRT* optimizes the path cost based on random sampling to ensure smooth trajectory and minimum energy consumption.

8. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S7 specifically includes: When executing the obstacle avoidance path, the feedback torque of each joint is collected, and a compensation current is generated by the PID controller to suppress overshoot; the compensation current ΔI = K_p · e + K_i \int e dt, where e is the deviation between the theoretical torque and the feedback torque, to prevent secondary collisions caused by inertia during obstacle avoidance.

9. The industrial robot anti-collision early warning method according to claim 1, characterized in that, Step S8 specifically includes: Activate the corresponding device based on the warning level: Level 1: Activate the robot's warning lights; Level 2: Send deceleration logs to the MES system; Level 3: Triggers the safety relay to cut off the drive power; Hardware response latency is controlled within 20ms, meeting the SIL 2 safety level; Step S9 specifically includes: recording the collision probability, actual distance deviation, and response time for each warning; updating the weight parameters of the LSTM network; and optimizing the prediction model based on the PPO algorithm of reinforcement learning to reduce model drift caused by environmental changes. Step S10 specifically includes: storing the evaluation results of step 9 into the graph database Neo4j, establishing "scene-action-effect" association rules; when a similar scene reappears, directly calling the historical best strategy, shortening the planning time to 100ms.

10. An industrial robot collision avoidance and early warning system, characterized in that, Use any one of the collision avoidance warning methods for industrial robots as described in claims 1 to 9.

Images