Path Planning for Multi-Arm Manipulators using Soft Actor-Critic with Hindsight Experience Replay

Matlab workspace simulation of SAC, TD3 and PRM algorithms using two 3-DOF open manipulators
- (For details of the open manipulator, see http://www.robotis.us/openmanipulator-x/)

In Matlab workspace simulation, SAC with HER, TD3 with HER, and PRM generate the optimal collision-free movement.

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.

Matlab simulation in configuration space for SAC, TD3 and PRM algorithm using two 3-DOF open manipulators

These are generated collision-free paths in configuration space when arbitrary starting points and goal points are given.

To be specific, the optimal path of the proposed method is shorter by 9.66% than that of PRM on average.

Compared with the TD3 with HER, on average, the paths by the proposed method are 7.6% shorter than those by TD3 with HER.

Experiment the operation based on the paths by SAC with HER in two 3-DOF open manipulator using ROS

The testing of the real manipulator's operation using two real 3-DOF open manipulators