Trajectory Planning of Robotic Manipulator in Dynamic Environment Exploiting Deep Reinforcement Learning

This study is about the implementation of a reinforcement learning algorithm in the trajectory planning of manipulators. We have a 7-DOF robotic arm to pick & place the randomly placed block at a random target point in an unknown environment. The obstacle is randomly moving which creates a hurdle in picking the object. The objective of the robot is to avoid the obstacle and pick the block with constraints to a fixed timestamp. In this literature, we have applied a deep deterministic policy gradient (DDPG) algorithm and compared the models’ efficiency with dense and sparse rewards.

I. Introduction

In this modern era, robots are commonly used in many applications such as picking and placing objects, welding, surgical, agricultural sectors, and many more. Industrial robots operate in complex environments where uncertainties and causalities may happen. In some applications, humans have to physically interact with the robot which is commonly known as Human-Robot Interaction (HRI). However, this interaction can raise the risk of safety concerns for humans and the environment.