CN-121973221-A - Multi-robot collaborative path planning and anti-collision dynamic obstacle avoidance welding method and system

CN121973221ACN 121973221 ACN121973221 ACN 121973221ACN-121973221-A

Abstract

The invention relates to the technical field of intelligent manufacturing and industrial robots, in particular to a multi-robot collaborative path planning and anti-collision dynamic obstacle avoidance welding method and a system, wherein the system integrates micro edge computing modules at each robot joint driver, triggers millisecond-level local decisions based on the change rate of health states, through a four-stage closed loop flow, the accurate identification and the quick response of a high curvature path, neighbor approximation, arc interference and joint degradation concurrent scene are realized through health dynamic monitoring, composite risk assessment, health coupling obstacle avoidance execution and self-adaptive evolution. The invention solves or at least reduces the problem of obstacle avoidance failure under composite faults caused by health monitoring and obstacle avoidance logic fracture and centralized architecture response hysteresis in the prior art, and provides a multi-robot collaborative path planning and anti-collision dynamic obstacle avoidance welding method and system.

Inventors

WANG MIN
DOU JINXING
HUANG TAO
YUAN CHAO
HU XUEJUN
Yan Shuangliang
ZHANG XIAOPING
WANG LIANG
LIU KAI
BAN WEI
Shang Aoxiang
HE ZHENPENG

Assignees

中交第二航务工程局有限公司

Dates

Publication Date: 20260505
Application Date: 20260209

Claims (10)

1. A multi-robot collaborative path planning and anti-collision dynamic obstacle avoidance welding method is characterized in that a miniature edge computing module is deployed in the interior of a shell of each joint driver of a robot, and is connected with signal output ends of a joint encoder, a temperature sensor and a motor driver through a hardware signal splitter, wherein the method comprises the following steps: the method comprises the following steps of (S1) dynamically monitoring the health state, namely synchronously collecting the vibration waveform, the temperature rise rate and the load torque change rate of the joint, and triggering a path freezing instruction and starting a joint-crossing data channel when the vibration and temperature rise change rate continuously exceeds a preset threshold value after anti-interference filtering; (S2) composite scene risk assessment, namely judging a composite risk scene by combining local space curvature and obstacle distance, marking risk levels according to multi-parameter mutation combination, expanding a safety margin threshold value under the highest risk level, broadcasting a cooperative obstacle avoidance instruction through a wireless beacon, and executing an avoidance action by the adjacent robot in response to the instruction; The method comprises the following steps of (S3) executing healthy coupling obstacle avoidance, namely limiting joint speed and verifying structural stability based on adjacent joint data under the highest risk level, compensating vector offset of a path according to safety margin, ensuring reliability of the corrected path through a multi-layer verification mechanism, and gradually recovering speed and closing a cross channel according to a healthy state; and (S4) self-adaptive evolution, namely recording event key parameters into a health library, optimizing safety margin mapping logic through a neural network model when the data volume reaches a threshold value, dynamically adjusting a cooperative waiting period according to the response efficiency of the neighbor machine, and self-adaptively updating a health mutation threshold value based on the historical trigger frequency.
2. The method of claim 1, wherein the path freeze command in S1 is sent via a high priority bus protocol to interrupt a path update command from the central controller while activating a visual alert signal at the end of the gun.
3. The method of claim 1, wherein the cooperative obstacle avoidance instruction in S2 includes a hierarchical code, and wherein a highest risk code triggers a neighbor to fall back a predetermined distance along a predetermined security direction and return a confirmation signal.
4. The method of claim 1, wherein the path compensation in S3 comprises: Constructing a virtual reference system fusing the pose of adjacent joints to offset track drift; Inserting a smooth transition section when the deviation between the original path and the corrected path exceeds the limit; the waypoint coordinate modulation is dynamically adjusted based on real-time health characteristics.
5. The method of claim 5, wherein the validation of the defrost request is required to satisfy a double condition: transmitting a request frame with a state lock to a central controller; and switching the visual signal of the tail end of the welding gun to a preset safety state.
6. The method of claim 1, wherein the speed recovery in S3 is divided into a multi-stage progressive ramp up, each stage requiring a decreasing healthy mutation rate threshold to be met, otherwise reverting to the previous stage.
7. The method of claim 1, wherein the neural network model in S4 takes as input a historical mutation rate and a safety margin, and outputs a nonlinear adjustment map instead of a fixed ratio logic.
8. The method of claim 1, wherein the adaptive updating of the health mutation threshold in S4 comprises: When the trigger frequency of the continuous tasks is too high, the threshold value is adjusted upwards; the threshold is adjusted down when the continuous task is not triggered; The new threshold takes effect after the system is restarted and records the change log.
9. A system for implementing the method of any one of claims 1-8, comprising: a plurality of welding robot units, each unit joint driver is embedded with a miniature edge calculation module; the module is connected with the encoder, the temperature sensor and the motor driver through the hardware branching unit; The module internal circuit is configured to execute the whole flow of S1 to S4, and realize hardware level coupling of the health state and obstacle avoidance control.
10. The system of claim 9, wherein the micro-edge computing module is integrated with the motor driver housing via a thermally conductive medium, with integrated analog-to-digital conversion, hardware filtering, and multi-unit processing circuitry.

Description

Multi-robot collaborative path planning and anti-collision dynamic obstacle avoidance welding method and system Technical Field The invention belongs to the technical field of intelligent manufacturing and industrial robots, and particularly relates to a multi-robot collaborative path planning and anti-collision dynamic obstacle avoidance welding method and system. Background Under the trend of intelligent manufacturing towards high precision and high reliability, a multi-robot cooperative welding system has become a key technical support for realizing integrated forming of complex components in the high-end manufacturing fields such as medical equipment, aerospace and the like. The application scene has very strict requirements on operation continuity, path precision and system safety, and not only is the robot required to complete millimeter-level positioning welding track tracking in a narrow space, but also absolute anti-collision capability is required to be maintained in a high-constraint environment with frequent intervention of dynamic obstacles and rapid working condition switching. Under the background, path planning and dynamic obstacle avoidance are not only space avoidance problems at the geometric level, but are evolved into multidimensional coupling control problems fusing body health states, environment awareness and collaborative decisions. In the prior art, a patent with publication number CN116619381B proposes a high-dimensional path planning method based on bidirectional node tree expansion, and the efficiency and stability of collaborative path generation are obviously improved under static or preset obstacle conditions for a gantry type double-welder robot system. By constructing a feasible path tree in the global configuration space, the motion interference between mechanical arms can be effectively avoided, and the engineering applicability is good in early welding scenes mainly comprising fixed tools and low dynamic interference. However, as the welding task progresses toward high curvature structures (curvature >25 °/m) and miniaturized components, this approach exposes fundamental limitations. The central controller centralized decision architecture on which it depends, the path update period is limited by the communication link delay (more than or equal to 1200 ms), and there is no way to respond to burst disturbance in millisecond level at all. More importantly, the obstacle avoidance logic and the real-time running state of the robot body are completely disjointed, and health parameters such as joint vibration, temperature rise and the like are only taken as the basis of offline maintenance and are not embedded into an obstacle avoidance triggering mechanism. If lubrication failure and high-load welding occur simultaneously, positioning drift (the measured value can reach 1.8 mm) is caused, and the system is particularly easy to generate unexpected collision due to lack of perception of hidden faults. Still another representative scheme CN115202365B, which combines an improved swarm algorithm with three-dimensional grid modeling, optimizes the global path by introducing the torch rotation angle influence factor, and provides a certain improvement in dynamic obstacle avoidance. However, its core is still built on a centralized software processing paradigm. All sensor data is uploaded to the main control unit for fusion calculation, which results in the loss of local response capability. Moreover, the risk determination depends on a fixed threshold, the working condition self-adaptability is lacked during the conventional welding and the precise welding mode switching, and the false triggering rate exceeds 35% in a high-sensitivity scene. It is particularly outstanding that this solution does not incorporate the dynamic evolution of the health state of the joint into the obstacle avoidance decision closed loop, and there is no way to identify soft collision risks caused by progressive degradation of bearing wear, overheating, etc. Such risks, although not immediately leading to downtime, can continuously accumulate positioning errors, eventually causing irreversible quality defects at the critical welds, and even equipment damage. A common drawback of these solutions is that the logic of health monitoring and obstacle avoidance execution is split in the architectural design. The traditional system takes health data as post-diagnosis information and takes obstacle avoidance as pure geometric space problem, but ignores abrupt changes of joint health, such as abrupt increase of vibration waveform change rate and abnormal temperature rise rate, which are pre-arranged collision symptoms in a high-precision welding strong coupling scene. When multiple factors such as health degradation, high curvature paths, adjacent machine approximation, arc interference and the like are overlapped to form a quadruple composite fault, the characteristics of communication bottleneck and threshold stif