10 lessons · 10–16 h (Parts 1–2)

Stage 14 · Advanced Robotics / Reinforcement Learning

Parts 1–2: RL foundations for robots (MDP, policy/value, exploration, reward shaping) plus value-based deep RL / DQN intuition (why tables break, function approximation, Q-net + replay + target, stability traps). Soft-recommend Stages 11–13; hard locks cleared; no payment gating. Browser educational toys only — NOT MuJoCo/Isaac/stable-baselines. Parts 3–4 (policy gradients, sim-to-real caveats) planned later.

• 1. Why Robots Learn
• 2. States, Actions & Rewards
• 3. Policies & Value
• 4. Episodes, Exploration vs Exploitation
• 5. Reward Shaping for Manipulation
• 6. Why Tabular Q Breaks for Robots
• 7. Function Approximation Intuition
• 8. DQN Core Ideas
• 9. Stability Traps & Overestimation
• 10. DQN Intuition Challenge