A collection of reinforcement learning and imitation learning projects for robotic control tasks.

• Imitation learning for 2D visual servoing

• Contains dataset handling and simple CNN policy

• Includes synthetic data generation utilities

• Soft Actor-Critic implementation for servo control

• Handles both image and proprioceptive (qpos) inputs

• GPU-optimized replay buffer implementation

• Full physics environment for Unitree Go2 robot
• Implements reward functions and PD control
• Supports parallel environments for efficient training

• PyBullet-based environment for object fixation task
• Discrete action space for camera control
• Includes memory debugging utilities

• Sequence prediction policy for multi-step actions

• Temporal convolutional network architecture

• PyBullet deployment integration

• Mouse movement imitation from screen capture

• Temporal CNN with spatial pyramid pooling

• Normalized coordinate prediction

pip install -r requirements.txt # Core dependencies: # - PyTorch # - PyBullet # - h5py # - stable-baselines3 # - genesis-sim