Mechanical arm anti-collision analysis system based on explosive ordnance disposal robot

Abstract

The invention discloses a mechanical arm anti-collision analysis system based on an explosive ordnance disposal robot, and relates to the technical field of public safety, the mechanical arm anti-collision analysis system comprises a multi-source heterogeneous data acquisition module, a space-time alignment characteristic distillation module, an attribute-track coupling prediction module and the like, and precise anti-collision is achieved by fusing obstacle attributes and tracks and gradient protection boundaries. The system has the advantages that attribute data such as obstacle material components and mechanical arm state data are synchronously obtained through the multi-source heterogeneous data acquisition module, features are extracted through fusion of the space-time alignment and feature distillation module, and a standardized data set containing rigid, explosive and quality tags is generated; and the attribute tag is incorporated into a state space by the obstacle attribute-track coupling prediction model, and the obstacle attribute and the motion track are associated to carry out collaborative analysis, so that the defect of separation of dynamic obstacle attribute and track analysis in the prior art is effectively overcome, and an accurate attribute-track association basis is provided for anti-collision analysis.

Description

Technical Field

[0001] This invention relates to the field of public safety technology, specifically to a collision avoidance analysis system for a robotic arm based on a bomb disposal robot. Background Technology

[0002] The high risk of explosives means that direct human operation carries a fatal risk. As a result, bomb disposal robots have become the core equipment to replace humans in completing high-risk bomb disposal tasks. Among them, the robotic arm, as the key execution component of the bomb disposal robot, undertakes core operations such as grabbing, transferring and dismantling explosives. The safety and accuracy of its operation directly determine the success or failure of the bomb disposal task, while collision avoidance analysis is the core technical support to ensure the safe operation of the robotic arm.

[0003] Existing collision avoidance technologies for bomb disposal robots mainly rely on a single sensor to collect environmental data and combine it with a fixed safety threshold to achieve collision warning. This approach has certain drawbacks. First, it is difficult to simultaneously acquire dynamic obstacle material composition, rigidity, and other attribute data with the robot's state data. It also cannot correlate obstacle attributes with motion trajectories for collaborative analysis, resulting in a lack of accurate basis for collision avoidance analysis. Second, fixed protection thresholds are difficult to adapt to bomb disposal scenarios of varying complexity, and there is a lack of visualized risk information and avoidance strategy recommendations, leading to low human-machine decision-making efficiency. To address these issues, we propose a collision avoidance analysis system for a bomb disposal robot's robotic arm. Summary of the Invention

[0004] The purpose of this invention is to provide a collision avoidance analysis system for a robotic arm based on a bomb disposal robot.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a collision avoidance analysis system for a robotic arm based on an explosive ordnance disposal robot, comprising a collision avoidance analysis system, wherein the collision avoidance analysis system includes the following modules:

[0006] Multi-source heterogeneous data acquisition module: It adopts LiDAR + millimeter-wave radar + hyperspectral camera array to collect obstacle point cloud, velocity vector and material composition data, and simultaneously collects robotic arm joint angle, end position and load data. It avoids timing deviation through timestamp calibration. The multi-source heterogeneous data acquisition module transmits the collected bomb disposal site environmental data and robotic arm motion status data to the spatiotemporal alignment and feature distillation module.

[0007] Spatiotemporal alignment and feature distillation module: Employs dynamic time warping to align multi-sensor data, extracts feature information through Transformer cross-modal attention mechanism, associates contour and material composition features, and generates a standardized dataset with rigidity, explosiveness and mass labels. After spatiotemporal calibration and feature fusion of the data, the spatiotemporal alignment and feature distillation module generates a standardized dataset with obstacle attribute labels, which is then transmitted to the obstacle attribute-trajectory coupling prediction module.

[0008] Obstacle attribute-trajectory coupled prediction module: Based on the attribute-trajectory coupled prediction model, the obstacle attribute label and trajectory uncertainty are incorporated into the state space, and the obstacle motion trajectory, confidence level and collision risk level are output. The obstacle attribute-trajectory coupled prediction module outputs the obstacle motion trajectory prediction results and the preliminary collision risk assessment results, which are respectively transmitted to the dynamic protection boundary generation module and the human-machine collaborative decision-making interaction module.

[0009] Dynamic protection boundary generation module: Based on obstacle attributes and speed, a gradient safety distance mechanism is constructed to generate differentiated protection boundaries, balancing protection redundancy and operation efficiency. The dynamic protection boundary generation module generates adaptive protection boundary signals based on the prediction results and transmits them to the robotic arm motion control adaptation module.

[0010] Robotic arm motion control adapter module: Receives protection signals and decision signals, combines the dynamic characteristics of the robotic arm and operational requirements to generate anti-collision control commands and verify their feasibility;

[0011] Human-machine collaborative decision-making interaction module: Visualizes risk information, retains human control and records decision differences. The human-machine collaborative decision-making interaction module receives human decision-making instructions from operators or outputs strategy recommendation results, and transmits them to the robotic arm motion control adaptation module.

[0012] Data storage and backtracking module: Stores data from each module and supports querying and replaying by timestamp and task number, providing data support for review and optimization. The multi-source heterogeneous data acquisition module, spatiotemporal alignment and feature distillation module, obstacle attribute-trajectory coupling prediction module, dynamic protection boundary generation module, robotic arm motion control adaptation module, and human-machine collaborative decision-making interaction module all transmit their respective input and output data to the data storage and backtracking module in real time.

[0013] As a further aspect of the present invention: in the multi-source heterogeneous data acquisition module, the lidar sampling frequency is ≥100Hz and the spatial resolution is ≥0.05m, the millimeter-wave radar speed accuracy is ≥0.1m / s, the hyperspectral camera spectral resolution is ≥5nm and the time synchronization error is ≤10ms, the robotic arm data is acquired through the CAN bus, and the joint angle accuracy is ±0.01° and the load accuracy is ±0.05kg.

[0014] As a further aspect of the present invention: in the spatiotemporal alignment and feature distillation module, the cross-modal attention mechanism adjusts the weights according to the feature information entropy, and the attribute label generation rules are as follows: surface curvature > 5° / m is high rigidity, < 2° / m is flexible, hyperspectral matching similarity > 85% is high explosive probability, 60%-85% is medium explosive probability, and < 60% is low explosive probability, and the quality label is calculated by point cloud volume and material density.

[0015] As a further aspect of the present invention: the state space of the obstacle attribute-trajectory coupling prediction module contains 10 types of parameters, 5 on the robotic arm side: joint angle, angular velocity, end effector position, velocity, and load; and 5 on the obstacle side: current position, velocity, rigidity coefficient, explosive probability, and mass, for a total of 28 dimensions. The reward function of the prediction model is set with dynamic coefficients. When the probability of explosion is greater than 70% =10, 30%-70% =5, <30% =2, joint angular acceleration change rate <5° / s 2 Bonus +0.5, >10° / s 2 Penalty -1.0, trajectory prediction period 1-3 seconds, confidence level calculated using Kalman filter covariance.

[0016] As a further aspect of the present invention: in the dynamic protection boundary generation module, the gradient safety distance is determined by the risk entropy value calculated using the following formula:

[0017] ;

[0018] in, For risk entropy value, The higher the value, the greater the required safe distance. This represents the probability distribution of each attribute category of the obstacle. Represents rigid probability, Indicates the probability of explosion. Represents a flexible probability, and + + , , These are the attribute and the speed weight coefficient, respectively. Take 0.6-0.8, Take a value of 0.2-0.4. For the speed of the obstacle, This is a scene correction factor, with a value ranging from 0.8 to 1.5;

[0019] The formula for calculating safe distance is as follows:

[0020] ;

[0021] in, For the final safe distance, Based on the basic safety distance, the end diameter is >0.1. hour ≤0.1 hour ;

[0022] For highly explosive targets with a radius greater than 70%, a double-layer protective ring is used: core radius (Hand grenade 0.3m, explosive charge 0.5m, artillery shell 0.8m), buffer radius ,satisfy ≤ ;

[0023] Triggering rule: Distance between the robotic arm and the obstacle Reduce speed to 50%. Emergency stop and ≥ Resume operations.

[0024] As a further aspect of the present invention: the instruction generation process of the robotic arm motion control adaptation module is as follows: receiving signals and reading motion status and task parameters; deceleration command reducing joint angular velocity by 50% and torque ≤ 80% of rated torque; stop command stopping smoothly within 0.5 seconds; path adjustment command adopting a path planning mechanism adapted for bomb disposal scenarios, with an additional 0.1m high-risk obstacle avoidance margin for easily explosive obstacles; and dynamic verification of the generated motion parameters to ensure that joint torque does not exceed 90% of rated torque and angular acceleration does not exceed 20° / s². 2 .

[0025] As a further aspect of the present invention: the human-machine collaborative decision-making interaction module has a visual interface: a top view that marks the position and trajectory of the robotic arm / obstacle, color coding of risk (red = high explosiveness, yellow = high rigidity, blue = low risk), a details bar that displays attributes, collision time and confidence level, human command response ≤200ms, pop-up prompts when deviation >0.2m, and data backtracking of difference data.

[0026] As a further aspect of the present invention: the data storage and backtracking module adopts a distributed database to store raw data, intermediate results (datasets, trajectory predictions, risk entropy values) and control interaction data. The backtracking function supports filtering by timestamp, task number and event type, playback at 1-10x speed and synchronous display of module data, and automatically generates a review report containing the number of warnings, avoidance success rate, manual intervention rate and fault records.

[0027] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0028] 1. This invention synchronously acquires attribute data such as the material composition of obstacles and the state data of the robotic arm through a multi-source heterogeneous data acquisition module. After spatiotemporal alignment and feature distillation modules, features are extracted and a standardized dataset containing rigidity, explosiveness and mass labels is generated. Then, the attribute labels are incorporated into the state space by the obstacle attribute-trajectory coupling prediction model, and the obstacle attributes and motion trajectory are correlated to carry out collaborative analysis. This effectively solves the defect of the separation between dynamic obstacle attributes and trajectory analysis in the prior art, and provides accurate attribute-trajectory correlation basis for collision avoidance analysis.

[0029] 2. This invention uses a dynamic protection boundary generation module to replace a fixed threshold with a gradient safety distance mechanism, and combines risk entropy values ​​to generate differentiated protection boundaries, solving the problem that protection thresholds are not suitable for bomb disposal scenarios. At the same time, relying on the human-machine collaborative decision-making interaction module, risk information is visualized, avoidance strategies are pushed, and human control is retained, which solves the defects of inefficient human-machine decision-making. Ultimately, it improves the flexibility and decision-making efficiency of the collision avoidance analysis of the bomb disposal robot arm and ensures the safety of bomb disposal operations. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the system flow in an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the attribute-trajectory coupling prediction process in an embodiment of the present invention. Detailed Implementation

[0032] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0033] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0034] Please see the appendix Figure 1 - Appendix Figure 2 The present invention discloses a collision avoidance analysis system for a robotic arm based on a bomb disposal robot, comprising a collision avoidance analysis system, which includes the following modules:

[0035] Multi-source heterogeneous data acquisition module: It adopts LiDAR + millimeter-wave radar + hyperspectral camera array to collect obstacle point cloud, velocity vector and material composition data, and simultaneously collects robotic arm joint angle, end position and load data, and avoids timing deviation through timestamp calibration;

[0036] Spatiotemporal alignment and feature distillation module: Employs dynamic temporal warping to align multi-sensor data, extracts feature information through the Transformer cross-modal attention mechanism, associates contour and material composition features, and generates a standardized dataset with rigidity, explosiveness and mass labels;

[0037] Obstacle attribute-trajectory coupled prediction module: Based on the attribute-trajectory coupled prediction model, obstacle attribute labels and trajectory uncertainties are incorporated into the state space, and the obstacle motion trajectory, confidence level and collision risk level are output;

[0038] Dynamic protection boundary generation module: Based on obstacle attributes and speed, a gradient safety distance mechanism is constructed to generate differentiated protection boundaries, balancing protection redundancy and operational efficiency;

[0039] Robotic arm motion control adapter module: Receives protection signals and decision signals, combines the dynamic characteristics of the robotic arm and operational requirements to generate anti-collision control commands and verify their feasibility;

[0040] Human-machine collaborative decision-making interaction module: Visualizes risk information, retains human control, and records decision differences;

[0041] Data storage and backtracking module: Stores data from each module, supports querying and replaying by timestamp and task number, and provides data support for review and optimization.

[0042] In one embodiment of the present invention: in the multi-source heterogeneous data acquisition module, the lidar sampling frequency is ≥100Hz and the spatial resolution is ≥0.05m, the millimeter-wave radar speed accuracy is ≥0.1m / s, the hyperspectral camera spectral resolution is ≥5nm and the time synchronization error is ≤10ms, the robotic arm data is acquired through the CAN bus, and the joint angle accuracy is ±0.01° and the load accuracy is ±0.05kg.

[0043] In one embodiment of the present invention: in the spatiotemporal alignment and feature distillation module, the cross-modal attention mechanism adjusts the weights according to the feature information entropy, and the attribute label generation rules are: surface curvature > 5° / m is high rigidity, < 2° / m is flexible, hyperspectral matching similarity > 85% is high explosive probability, 60%-85% is medium explosive probability, and < 60% is low explosive probability, and the quality label is calculated by point cloud volume and material density.

[0044] In one embodiment of the present invention: the state space of the obstacle attribute-trajectory coupling prediction module contains 10 types of parameters, 5 on the robotic arm side: joint angle, angular velocity, end effector position, velocity, and load; and 5 on the obstacle side: current position, velocity, rigidity coefficient, explosive probability, and mass, for a total of 28 dimensions. The reward function of the prediction model is set with dynamic coefficients. When the probability of explosion is greater than 70% =10, 30%-70% =5, <30% =2, joint angular acceleration change rate <5° / s 2 Bonus +0.5, >10° / s 2 Penalty -1.0, trajectory prediction period 1-3 seconds, confidence level calculated using Kalman filter covariance.

[0045] In one embodiment of the present invention: in the dynamic protection boundary generation module, the gradient safety distance is determined by the risk entropy value calculated using the following formula:

[0046] ;

[0047] in, For risk entropy value, The higher the value, the greater the required safe distance. This represents the probability distribution of each attribute category of the obstacle. Represents rigid probability, Indicates the probability of explosion. Represents a flexible probability, and + + , , These are the attribute and the speed weight coefficient, respectively. Take 0.6-0.8, Take a value of 0.2-0.4. For the speed of the obstacle, This is a scene correction factor, with a value ranging from 0.8 to 1.5;

[0048] The formula for calculating safe distance is as follows:

[0049] ;

[0050] in, For the final safe distance, Based on the basic safety distance, the end diameter is >0.1. hour ≤0.1 hour ;

[0051] For highly explosive targets with a radius greater than 70%, a double-layer protective ring is used: core radius (Hand grenade 0.3m, explosive charge 0.5m, artillery shell 0.8m), buffer radius ,satisfy ≤ ;

[0052] Triggering rule: Distance between the robotic arm and the obstacle Reduce speed to 50%. Emergency stop and ≥ Resume operations.

[0053] In one embodiment of the present invention, the command generation process of the robotic arm motion control adaptation module is as follows: receiving signals and reading motion status and task parameters; deceleration command reducing joint angular velocity by 50% and torque ≤ 80% of rated torque; stop command stopping smoothly within 0.5 seconds; path adjustment command adopting a path planning mechanism adapted for bomb disposal scenarios, setting an additional 0.1m high-risk obstacle avoidance margin for easily explosive obstacles; and performing dynamic verification on the generated motion parameters to ensure that joint torque does not exceed 90% of rated torque and angular acceleration does not exceed 20° / s². 2 .

[0054] In one embodiment of the present invention: the human-machine collaborative decision-making interaction module has a visual interface: the top view marks the position and trajectory of the robotic arm / obstacle, the risk is color-coded (red = high explosiveness, yellow = high rigidity, blue = low risk), the details bar displays the attributes, collision time and confidence level, the human command response is ≤200ms, the pop-up prompt is when the deviation is >0.2m, and the difference data is stored and traced back.

[0055] Example 1, please refer to the appendix. Figure 1 - Appendix Figure 2 Bomb disposal operations in an open outdoor setting, targeting suspected hand grenades:

[0056] Applied to a bomb disposal scenario in an urban square, with the target being a suspected hand grenade (a highly explosive obstacle), the system's module parameters and operating procedures are as follows:

[0057] Multi-source heterogeneous data acquisition module: LiDAR acquires grenade point cloud at 120Hz sampling frequency and 0.04m spatial resolution; millimeter-wave radar detects grenade stationary (velocity 0m / s) with an accuracy of 0.05m / s; hyperspectral camera matches nitro compound features at 5nm spectral resolution with a similarity of 92%; synchronously acquires joint angles (accuracy ±0.008°) and end-effector load (0.8kg) of robotic arm (6 joints) with a time synchronization error of 8ms.

[0058] Spatiotemporal alignment and feature distillation module: Align data with DTW, assign a weight of 0.8 to hyperspectral features (high information entropy) using cross-modal attention mechanism, and generate a labeled dataset of "high rigidity (surface curvature 6° / m), high explosive probability (92%), and mass 0.8kg".

[0059] Obstacle attribute-trajectory coupling prediction module: The state space contains 28 parameters such as robot arm joint angles (6D) and angular velocity (6D), and the reward function... =10 (probability of explosion 92% > 70%), the grenade is predicted to stop within 1.5 seconds (trajectory confidence 95%, covariance 0.008m²), and the collision risk level is "high".

[0060] Dynamic protection boundary generation module: Take =0.7、 =0.3、 =0.9, =0.9、 =0.92、 =0.18 (sum after normalization is 1), calculate , Core radius =0.3 Buffer radius .

[0061] Robotic arm control and human-machine collaboration: The robotic arm approaches to within 0.35 meters. Reduce speed to 50% (original speed 0.2). / s→0.1 / s), torque ≤80% of rated value (rated 50N·m / s), →≤40N・ The human-machine interface pushes three sets of strategies, and the operator selects "optimal safety redundancy". The command response time is 150ms.

[0062] Data storage: Stores all raw data and intermediate results, and supports playback by task number "20240510-Square Explosive Ordnance Disposal 001".

[0063] Example 2, please refer to the appendix. Figure 1 - Appendix Figure 2 Bomb disposal operations in confined indoor spaces (targeting suspected explosive charges and flexible obstacles):

[0064] When applied to an indoor bomb disposal scenario, with the target being a suspected explosive charge (highly explosive) and a piece of fluttering fabric (a flexible barrier), the system operates as follows:

[0065] Data acquisition: LiDAR 100Hz sampling, 0.05m resolution; millimeter-wave radar detection speed of explosive charge 0.1m / s (accuracy 0.08m / s), material placement speed 0.3m / s; hyperspectral matching of explosive charge similarity 88%, material placement similarity 35%; robotic arm (8 joints) data acquisition via CAN bus, joint angular velocity accuracy ±0.09° / s, load 1.2kg, synchronization error 10ms.

[0066] Feature distillation: Cross-modal attention assigns a weight of 0.7 to the lidar contour features (with many indoor obstructions), generating labels for explosive charges "high rigidity (curvature 7° / m), high explosiveness (88%), mass 1.2kg" and fabric "flexible (curvature 1.5° / m), low explosiveness (35%), mass 0.1kg".

[0067] Trajectory prediction: Predict the trajectory of the explosive charge in 2 seconds (confidence level 93%, covariance 0.009m²). =10, Fabric =2, angular acceleration rate of change 4° / s 2 Receive a +0.5 reward; risk level is "low".

[0068] Protective boundary: Take =0.8、 =0.2、 =1.4 (narrow scene), explosive charge (0.8×(0.95 2 +0.88 2 +0.17 2 () + 0.2 × 0.1) × 1.4 ≈ 1.9385, =0.3×(1+1.9385)=0.88155m, =0.5m, =0.5×(1+0.5×0.1)=0.525m, fabric =0.2×(1+0.4)=0.28m (simplified calculation for low-risk areas).

[0069] Control and coordination: When the robotic arm avoids the explosive charge, a high-risk margin of 0.1m is set, but when it avoids the cloth, there is no margin. The interface prompts "Prioritize avoiding the explosive charge". Manual commands with a deviation of 0.15m (<0.2m) are executed directly without pop-up windows.

[0070] Data storage: Records the obstacle differentiation and processing process, and supports filtering and playback by "collision warning event".

[0071] This system addresses the disconnect between dynamic obstacle attributes and trajectory analysis by employing multi-source heterogeneous data acquisition, spatiotemporal alignment and feature distillation, and attribute-trajectory coupling prediction modules. Through a gradient-based safety distance mechanism and a human-machine collaborative decision-making module, it solves the problems of protection thresholds not matching bomb disposal scenarios and inefficient human-machine decision-making. Ultimately, it improves the accuracy and efficiency of collision avoidance and ensures the safety of bomb disposal operations in different scenarios.

[0072] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A collision avoidance analysis system for a robotic arm based on a bomb disposal robot, comprising a collision avoidance analysis system, characterized in that, The collision avoidance analysis system includes the following modules: Multi-source heterogeneous data acquisition module: It adopts LiDAR + millimeter-wave radar + hyperspectral camera array to collect obstacle point cloud, velocity vector and material composition data, and simultaneously collects robotic arm joint angle, end position and load data, and avoids timing deviation through timestamp calibration; Spatiotemporal alignment and feature distillation module: Employs dynamic temporal warping to align multi-sensor data, extracts feature information through the Transformer cross-modal attention mechanism, associates contour and material composition features, and generates a standardized dataset with rigidity, explosiveness and mass labels; Obstacle Attribute-Trajectory Coupled Prediction Module: Based on the obstacle attribute-trajectory coupled prediction model, obstacle attribute labels and trajectory uncertainties are incorporated into the state space, and the obstacle motion trajectory, confidence level and collision risk level are output. Dynamic protection boundary generation module: Based on obstacle attributes and speed, a gradient safety distance mechanism is constructed to generate differentiated protection boundaries, balancing protection redundancy and operational efficiency; Robotic arm motion control adapter module: Receives protection signals and decision signals, combines the dynamic characteristics of the robotic arm and operational requirements to generate anti-collision control commands and verify their feasibility; Human-machine collaborative decision-making interaction module: Visualizes risk information, retains human control, and records decision differences; Data storage and backtracking module: Stores data from each module, supports querying and replaying by timestamp and task number, and provides data support for review and optimization.

2. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that: In the multi-source heterogeneous data acquisition module, the lidar sampling frequency is ≥100Hz and the spatial resolution is ≥0.05m, the millimeter-wave radar speed accuracy is ≥0.1m / s, the hyperspectral camera spectral resolution is ≥5nm and the time synchronization error is ≤10ms, the robotic arm data is acquired through the CAN bus, and the joint angle accuracy is ±0.01° and the load accuracy is ±0.05kg.

3. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that: In the spatiotemporal alignment and feature distillation module, the cross-modal attention mechanism adjusts weights based on feature information entropy, and the attribute label generation rules are as follows: Surface curvature >5° / m indicates high rigidity, <2° / m indicates flexibility; hyperspectral matching similarity >85% indicates high explosive probability, 60%-85% indicates medium explosive probability, and <60% indicates low explosive probability; quality labels are calculated using point cloud volume and material density.

4. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that: The state space of the obstacle attribute-trajectory coupling prediction module contains 10 types of parameters: 5 types on the robotic arm side (joint angle, angular velocity, end effector position, velocity, and load) and 5 types on the obstacle side (current position, velocity, rigidity coefficient, explosive probability, and mass), for a total of 28 dimensions. The reward function of the obstacle attribute-trajectory coupling prediction model has dynamic coefficients. When the probability of explosion is greater than 70% =10, 30%-70% =5, <30% =2, joint angular acceleration change rate <5° / s 2 Bonus +0.5, >10° / s 2 Penalty -1.0, trajectory prediction period 1-3 seconds, confidence level calculated using Kalman filter covariance.

5. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that: In the dynamic protection boundary generation module, the gradient safety distance is determined by the risk entropy value calculated using the following formula: ； in, For risk entropy value, The higher the value, the greater the required safe distance. This represents the probability distribution of each attribute category of the obstacle. Represents rigid probability, Indicates the probability of explosion. Represents a flexible probability, and + + , , These are the attribute and the speed weight coefficient, respectively. For the speed of the obstacle, This is a scene correction factor; The formula for calculating safe distance is as follows: ； in, For the final safe distance, Based on the basic safety distance, the end diameter is >0.

1. hour End diameter ≤ 0.1 hour ; For highly explosive targets with a radius greater than 70%, a double-layer protective ring is used: core radius Buffer radius ,satisfy ≤ ; Triggering rule: Distance between the robotic arm and the obstacle Reduce speed to 50%. Emergency stop and ≥ Resume operations.

6. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that, The instruction generation process of the robotic arm motion control adaptation module is as follows: The system receives signals and reads motion status and task parameters. Deceleration commands reduce joint angular velocity by 50%, and torque is ≤80% of rated torque. Stop commands bring the system to a smooth stop within 0.5 seconds. Path adjustment commands employ a path planning mechanism adapted for bomb disposal scenarios, with an additional 0.1m high-risk obstacle avoidance margin for potentially explosive obstacles. The generated motion parameters are dynamically verified to ensure that joint torque does not exceed 90% of rated torque and angular acceleration does not exceed 20° / s². 2 .

7. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that, The human-machine collaborative decision-making interaction module is displayed through a visual interface, and the displayed content is as follows: The top view labels the position and trajectory of the robotic arm / obstacle, color-codes risks, the details bar displays attributes, collision time and confidence level, the human command response time is ≤200ms, a pop-up prompt is displayed when the deviation is >0.2m, and the difference data is stored and traced back.

8. The collision avoidance analysis system for a robotic arm based on a bomb disposal robot according to claim 1, characterized in that: The data storage and backtracking module uses a distributed database to store raw data, intermediate results, and control interaction data. The backtracking function supports filtering by timestamp, task number, and event type, and can play back data at 1-10x speed while simultaneously displaying the module data. It automatically generates a review report containing the number of warnings, avoidance success rate, manual intervention rate, and fault records.

Images