A reward function generation method using genetic algorithms: a robot soccer case study.
Çetin MeriçliTekin MeriçliH. Levent AkinPublished in: AAMAS (2010)
Keyphrases
- generation method
- robot soccer
- reward function
- case study
- reinforcement learning
- action selection
- markov decision processes
- reinforcement learning algorithms
- state space
- inverse reinforcement learning
- vision system
- transition probabilities
- multiple agents
- optimal policy
- multi robot
- temporal difference
- soccer game
- real robot
- real world
- transition model
- initially unknown
- real time
- random walk
- state variables
- link prediction
- markov chain
- search space
- multi agent