Approximate planning in POMDPs via MDP heuristic.
Yong LinXingjia LuFillia MakedonPublished in: IJCNN (2014)
Keyphrases
- partially observable markov decision processes
- partially observable
- markov decision processes
- planning under uncertainty
- approximate solutions
- state space
- point based value iteration
- dynamic programming
- markov decision problems
- reinforcement learning
- belief space
- classical planning
- forward search
- heuristic function
- optimal policy
- stochastic domains
- finite state
- sequential decision making problems
- belief state
- factored mdps
- heuristic search
- partial observability
- ai planning
- admissible heuristics
- planning problems
- decision theoretic planning
- decision problems
- dynamical systems
- decision theoretic
- continuous state
- dynamic programming algorithms
- partially observable markov decision process
- optimal solution
- predictive state representations
- probabilistic planning
- domain independent planning
- bayesian reinforcement learning
- markov decision process
- optimal planning
- initial state
- multi agent
- expected reward
- policy search
- planning process
- average reward
- planning graph
- reward function
- np hard
- domain independent
- linear program
- search algorithm
- simulated annealing
- linear programming
- markov chain
- policy iteration
- evaluation function
- dec pomdps
- optimal plans
- planning domains
- average cost