Swinging Search and Crawling Control

1. Our paper: A snake-inspired path planning algorithm based on reinforcement learning and self-motion for hyper-redundant manipulators.
2. A brief overview on my website.

Structure

1. arm_kine: my manipulator kinematics package.
   • forward kinematics: Given angles of the arm, calculate the location of the end.
   • Jacobian: Calculate the Jacobian matrix for inverse kinematics.
   • inverse kinematics: Given the velocity of the end, calculate the velocity of the angles.
2. coding_steps: steps 1-4 are runnable.
   • step1: Test the forward kinematics and the inverse kinematics. Control the end to move as desired.
   • step2: Control the end to move to the target position, and then start self-motion.
   • step3: Generate target end positions in the bucket for the task.
   • step4: Swinging search randomly, to find a collision-free configuration.
   • step5-9: Swinging search by DDPG. Some codes were left in a lab server and has been lost.
3. crawling: Crawling Control. All the videos and the pics in the paper can be reproduced by running these codes.

simple_16_2_speedup.mp4