DEAR: Disentangled Environment and Agent Representations for Reinforcement Learning without Reconstruction.
Ameya PoreRiccardo MuradoreDiego Dall'AlbaPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- multi agent
- learning agent
- dynamic environments
- exploration strategy
- multi agent environments
- autonomous agents
- agent receives
- robocup soccer
- agent learns
- initially unknown
- agent model
- agent environment
- software agents
- learning agents
- action selection
- state space
- intelligent agents
- multi agent systems
- current situation
- three dimensional
- decision making
- agent behavior
- mobile robot
- model free
- real time
- learning algorithm
- image reconstruction
- world model
- multiple agents
- reconstruction process
- partially observable
- learning capabilities
- reward function
- multiagent systems
- mobile agents
- machine learning
- markov decision process
- temporal difference
- autonomous learning
- state abstraction
- changing environment
- agent technology
- agent architecture
- markov decision processes