Training a Quadruped Robot to Walk with RL

Train a quadruped to walk with RL in Isaac Lab: 4,096 environments, terrain curriculum, domain randomization, teacher-student distillation, sim-to-real.

Behavior Cloning vs Reinforcement Learning: Robotics Approaches Compared

Behavior cloning vs reinforcement learning for robotics: covariate shift, DAgger, diffusion policies, AMP/GAIL, and today's practical BC-then-RL workflow.

Hindsight Experience Replay: Learning from Failed Attempts

Hindsight Experience Replay explained: relabel failed trajectories with achieved goals to learn from sparse rewards, HER strategies, SAC + HER training.

Isaac Gym Tutorial: NVIDIA's GPU-Accelerated Robot Simulator

Isaac Gym tutorial: GPU-accelerated robot simulation with thousands of parallel PyTorch environments, vectorized rewards, PPO, and domain randomization.

PyBullet Tutorial: Physics Simulation for Robot Learning

PyBullet tutorial for robot learning: GUI vs DIRECT modes, URDF loading, joint control, a Gymnasium reaching environment, PPO training, cameras, sim-to-real.

Building a Custom Gymnasium Environment from Scratch

Build a custom Gymnasium environment from scratch: action and observation spaces, reset and step, termination vs truncation, rewards, registration, and testing.

Train an AI to Play Pong from Pixels

Train an AI to play Pong from raw pixels with DQN: preprocessing, frame stacking, a CNN architecture, experience replay, epsilon scheduling, and tuning.

Train an AI to Play Flappy Bird with Reinforcement Learning

Train an AI to play Flappy Bird with DQN: environment setup, state representation, reward shaping, epsilon-greedy exploration, tuning, and evaluation.

Deep Q-Networks vs PPO: Which Algorithm for Which Game?

DQN or PPO for your game? A practical comparison of action spaces, sample efficiency, and stability, with clear rules for arcade, racing, and physics games.

Soft Actor-Critic (SAC) Tutorial for Continuous Control

Soft Actor-Critic (SAC) tutorial for continuous control: the maximum-entropy objective, twin critics, entropy tuning, hyperparameters, and SAC vs TD3 vs PPO.