MazeExplorer: A Customisable 3D Benchmark for Assessing Generalisation in Reinforcement Learning.
Luke HarriesSebastian LeeJaroslaw RzepeckiKatja HofmannSam DevlinPublished in: CoG (2019)
Keyphrases
- reinforcement learning
- function approximation
- multi agent
- reinforcement learning algorithms
- learning algorithm
- real world
- state space
- dynamic programming
- machine learning
- supervised learning
- reinforcement learning methods
- optimal control
- action selection
- temporal difference
- control problems
- policy search
- case study
- multi agent reinforcement learning
- stochastic approximation
- temporal difference learning
- markov decision process
- real robot
- database
- learning classifier systems
- transfer learning
- optimal policy
- information retrieval
- artificial intelligence