Research on High-Precision Motion Planning of Large Multi-Arm Rock Drilling Robot Based on Multi-Strategy Sampling...

Area of Science:

• Robotics
• Artificial Intelligence
• Mechanical Engineering

Background:

• Optimal motion planning is critical for multi-arm robots in complex tasks like rock drilling.
• Existing algorithms face challenges in efficiency and accuracy for multi-arm systems.

Purpose of the Study:

• To develop a high-precision motion planning method for multi-arm rock drilling robots.
• To improve the adaptability and search efficiency of RRT* algorithms in dynamic environments.

Main Methods:

• Introduced Multi-Strategy Sampling RRT* (MSS-RRT*) incorporating DRL position sphere sampling, spatial random sampling, and goal-oriented sampling.
• Employed a dual Jacobi iterative inverse solution with forward kinematics error compensation for precise positioning.
• Utilized Hindsight Experience Replay-Obstacle Arm Transfer (HER-OAT) for training Deep Reinforcement Learning (DRL) models with the TD3 algorithm.
• Implemented a cylindrical bounding box method for collision avoidance between robot arms.

Main Results:

• Achieved a 94.15% improvement in motion planning accuracy compared to single Jacobi iteration.
• Reduced planning time to 20.71% of Informed-RRT* under optimal path conditions.
• Decreased path length by 21.58% compared to Quick-RRT* within the same time constraints.

Conclusions:

• The proposed MSS-RRT* method significantly enhances motion planning accuracy and efficiency for multi-arm rock drilling robots.
• The integration of multi-strategy sampling and advanced DRL techniques offers superior performance in complex robotic applications.
• This research provides a robust solution for optimizing path planning in multi-arm robotic systems, paving the way for more sophisticated autonomous operations.