Part 8 · Reinforcement learning for robots

Learning behavior from reward in simulation, and the design choices that make it work on robots.

Learn first

This part assumes the following. If any of it is new, learn it on AI Engineering from Scratch first:

• MDPs, States, Actions & Rewards
• Dynamic Programming — Policy Iteration & Value Iteration
• Temporal Difference — Q-Learning & SARSA
• Policy Gradient — REINFORCE from Scratch
• Actor-Critic — A2C and A3C
• Proximal Policy Optimization (PPO)
• Introduction to JAX

Lessons

#	Lesson	Run type	Time	Status
32	A MuJoCo RL environment	CPU	~45 min	✅ done
33	PPO for continuous control	CPU	~75 min	✅ done
34	Observation and action design	CPU	~60 min	✅ done
35	Reward design and reward hacking	CPU	~60 min	✅ done
36	Legged locomotion and massively parallel training	GPU	~90 min	✅ done