AI & Tooling · Open Source · Research · 2026
RL Drone Path Planning
PPO vs A* for 2D drone navigation in a custom Gymnasium environment — including a live dynamic-obstacle experiment.
PPO · Gymnasium · A* · Research · Drone · Autonomous Drones · AI ML
Overview
A custom Gymnasium environment models a drone in a 2D occupancy grid with range-sensor observations and shaped rewards. A PPO agent (Stable-Baselines3) is trained and compared against classical A* planning.
The dynamic-obstacle experiment shows where reinforcement learning genuinely beats classical planning — and where it doesn't.
Media
