Reinforcement Learning with Particle Swarm Optimization Policy (PSO-P) in Continuous State and Action Spaces.
Daniel HeinAlexander HentschelThomas A. RunklerSteffen UdluftPublished in: Int. J. Swarm Intell. Res. (2016)
Keyphrases
- continuous state and action spaces
- particle swarm optimization
- reinforcement learning
- action selection
- pso algorithm
- function approximators
- optimal policy
- continuous state
- policy search
- particle swarm optimization pso
- particle swarm optimization algorithm
- markov decision problems
- function approximation
- markov decision process
- inertia weight
- multi objective
- partially observable markov decision processes
- global optimization
- control policies
- differential evolution
- action space
- temporal difference
- partially observable
- policy gradient
- model free
- policy iteration
- swarm intelligence
- state space
- reward function
- ant colony optimization
- genetic algorithm
- state action
- control policy
- state dependent
- rl algorithms
- reinforcement learning algorithms
- convergence speed
- markov decision processes
- infinite horizon
- finite state
- approximate policy iteration
- average reward
- decision making
- machine learning
- finite horizon
- decision problems
- control strategies
- dynamic programming
- learning algorithm
- evolutionary computation
- learning tasks
- optimal control
- markov chain
- sparse approximation