Learning by Doing: Controlling a Dynamical System using Causality, Control, and Reinforcement Learning.
Sebastian WeichwaldSøren Wengel MogensenTabitha Edith LeeDominik BaumannOliver KroemerIsabelle GuyonSebastian TrimpeJonas PetersNiklas PfisterPublished in: NeurIPS (Competition and Demos) (2021)
Keyphrases
- dynamical systems
- reinforcement learning
- state space
- policy search methods
- reinforcement learning methods
- differential equations
- partially observable
- hidden state
- discrete event
- nonlinear dynamical systems
- dynamic systems
- optimal control
- nonlinear dynamics
- control system
- phase space
- function approximation
- linear quadratic
- adaptive control
- immune network
- predictive state representations
- lyapunov exponents
- machine learning
- linear systems
- control method
- connectionist networks
- action selection
- model free
- dynamical behavior
- past observations
- markov decision processes
- data streams