Policy Advisory Module for Exploration Hindrance Problem in Multi-agent Deep Reinforcement Learning.
Jiahao PengToshiharu SugawaraPublished in: PRIMA (2020)
Keyphrases
- reinforcement learning
- multi agent
- action selection
- optimal policy
- exploration exploitation tradeoff
- exploration strategy
- policy search
- exploration exploitation
- markov decision process
- function approximation
- control policies
- state space
- partially observable markov decision processes
- markov decision processes
- partially observable
- reinforcement learning algorithms
- actor critic
- action space
- approximate dynamic programming
- control policy
- model based reinforcement learning
- state action
- active exploration
- policy gradient
- control problems
- policy evaluation
- function approximators
- intelligent agents
- rl algorithms
- learning process
- multi agent environments
- markov decision problems
- state and action spaces
- partially observable environments
- model free
- policy iteration
- cooperative
- reward function
- temporal difference
- agent learns
- single agent
- long run
- reinforcement learning problems
- learning algorithm
- multiple agents
- decision problems
- state dependent
- machine learning
- multiagent systems
- reinforcement learning agents
- continuous state
- average reward
- partially observable markov decision process
- dynamic programming
- multi agent reinforcement learning
- learning agents
- multi agent systems
- transition model
- continuous state spaces
- learning agent
- transfer learning
- coalition formation
- finite state
- balancing exploration and exploitation