LC

Project article

Deep RL for Differential Drive Robot

Watercolor illustration for the reinforcement learning robot project

A from-scratch Arduino and Raspberry Pi robot controlled by a PPO policy for target reaching and obstacle avoidance.

This project was developed for the 2024 Reinforcement Learning class. We built a differential drive robot from scratch with Arduino and Raspberry Pi, then implemented a PPO-based policy to control it. The robot's objective was to reach a target while avoiding obstacles using only distance, angle-to-target, and distance-sensor observations.

Robot in action

Architecture developed for the robot

Architecture diagram showing the robot hardware and software components