This project involves applying reinforcement learning to a hexapod robot to learn how to walk toward a target position. The robot was built using Dynamixel AX-12A servos and controlled with a BeagleBone Blue board. A PyBullet simulation environment was created to train the robot using the Proximal Policy Optimization (PPO) algorithm implemented in PyTorch.
- Reinforcement learning for a hexapod robot
- PyBullet simulation environment
- PPO algorithm implementation in PyTorch
- Kalman Filter for state estimation
- Fine-tuning neural networks to adapt to real-world scenarios
- Python (PyBullet, PyTorch)
- BeagleBone Blue
- Dynamixel AX-12A Servos
- Reinforcement Learning (PPO)