Reinforcement Learning of Informed Initial Policies for Decentralized Planning.
Landon KraemerBikramjit BanerjeePublished in: ACM Trans. Auton. Adapt. Syst. (2014)
Keyphrases
- reinforcement learning
- macro actions
- optimal policy
- cooperative multi agent systems
- decentralized control
- partially observable markov decision processes
- multi agent
- markov decision problems
- goal oriented
- policy search
- markov decision processes
- action selection
- multiagent reinforcement learning
- partially observable
- stochastic domains
- function approximation
- markov decision process
- learning algorithm
- control policies
- fitted q iteration
- decentralized decision making
- reinforcement learning algorithms
- heuristic search
- state space
- reinforcement learning problems
- dec pomdps
- machine learning
- policy gradient methods
- cooperative
- dynamic programming
- distributed systems
- decision problems
- initial state
- peer to peer
- multi agent reinforcement learning
- continuous state
- partial observability
- hierarchical reinforcement learning
- control policy
- deterministic domains
- reward shaping
- predictive state representations
- optimal control
- model free
- temporal difference