REACT: Revealing Evolutionary Action Consequence Trajectories for Interpretable Reinforcement Learning.
Philipp AltmannCéline DavignonMaximilian ZornFabian RitzClaudia Linnhoff-PopienThomas GaborPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- action selection
- partially observable domains
- action space
- function approximation
- external events
- reward shaping
- state space
- view invariant
- state action
- genetic algorithm
- model free
- evolutionary computation
- moving object trajectories
- trajectory data
- multi agent
- reinforcement learning algorithms
- fitted q iteration
- learning algorithm
- human actions
- markov decision processes
- temporal difference
- robotic control
- moving objects
- dynamic programming
- transfer learning
- transition model
- neural network
- motion patterns
- agent receives
- genetic programming
- multi agent reinforcement learning
- continuous state
- reasoning about actions
- evolutionary optimization
- evaluation function
- current situation
- motion trajectories
- optimal control