Continuous State-Action-Observation POMDPs for Trajectory Planning with Bayesian Optimisation.
Philippe MorereRomán MarchantFabio RamosPublished in: IROS (2018)
Keyphrases
- continuous state
- trajectory planning
- action space
- continuous action
- reinforcement learning
- partially observable markov decision processes
- motion planning
- state action
- robot navigation
- state space
- obstacle avoidance
- robot manipulators
- finite state
- markov decision processes
- path planning
- dynamic environments
- policy search
- point based value iteration
- action selection
- state dependent
- real valued
- control policies
- planning problems
- evaluation function
- dynamical systems
- genetic algorithm
- single agent
- mobile robot
- optimal policy
- dynamic programming
- domain independent
- neural network
- stochastic processes
- average reward
- optimal solution
- bayesian networks
- computer vision
- optimal path
- machine learning