Two-Link Robotic Arm Control

Welcome 👋

Overview

1. Implement two control methods to guide a two-link robotic arm’s end effector to any specified 2D position.

2. In the first phase, I used an MPC strategy combined with a neural network to predict joint accelerations, optimizing the distance between the target and the end effector for efficient target reaching.

3. In the second phase, I applied model-free RL methods (DQN and PPO). DQN used a custom Q-network with experience replay, while PPO, implemented with Stable-Baselines3, leveraged parallel environments to speed up training, allowing the arm to learn target-reaching autonomously.