Deep Reinforcement Learning based Path Planning for Multi-Arm Manipulator with Periodically Moving Obstacles

Matlab dynamic environment simulation of SAC, TD3 and PRM algorithms for two 3-DOF open manipulators

In Matlab simulation, when arbitrary starting and goal points are given, SAC with HER, TD3 with HER, and PRM compute the optimal collision-free path.

The simulation video shows both the workspace and the configuration space of the robot.

Training is done using Tensorflow 2.x and testing (path generation and manipulator operation) is done in Matlab using the trained network.

The SAC with HER method leads to the smoothest and shortest movement than the other algorithms.

Experiment with the operation based on the paths by SAC with HER in two 3-DOF open manipulators using ROS in a dynamic environment

Link : Video 3

The testing of the real manipulator's operation using two real 3-DOF open manipulators (open-manipulator from robotis.com).

When arbitrary starting and goal points are given, the SAC agent computes the optimal collision-free path.

Learning is done using Tensorflow and testing experiment (path generation and manipulator operation) is done using the open-manipulator from Robotis. It is verified that the proposed method indeed generates the collision-free path for the real manipulator

with both static and periodically moving obstacles.