Reward machines — structured reward specification.
hardAnswer
- Represent complex non-Markovian rewards as finite-state automata: state tracks task progress (e.g., 'first find A, then find B').
- Enables decomposition + credit assignment across sub-goals.
- Alternative to hand-crafted shaping.
- Foundation of LTL (Linear Temporal Logic) constrained RL.
- Applications: robotics multi-step tasks, curriculum learning.
- Modern extension: neuro-symbolic RL combining reward machines with NNs.
Check yourself — multiple choice
- Random
- Represent non-Markovian rewards as FSA tracking task progress ('find A then B'); enables decomposition + credit assignment; foundation of LTL-constrained + neuro-symbolic RL
- Same as shaping
- Not real
Reward machines: FSA over sub-goals; LTL RL; neuro-symbolic extension.
#reward-design#theory
Practise Reinforcement Learning
214 interview questions in this topic.