Approximate Linear Programming for Constrained Partially Observable Markov Decision Processes.
Pascal PoupartAarti MalhotraPei PeiKee-Eung KimBongseok GohMichael BowlingPublished in: AAAI (2015)
Keyphrases
- linear programming
- incremental pruning
- partially observable markov decision processes
- dynamic programming
- point based value iteration
- exact solution
- approximate solutions
- finite state
- planning under uncertainty
- belief state
- linear program
- np hard
- optimal policy
- reinforcement learning
- decision problems
- partially observable stochastic games
- dynamical systems
- state space
- belief space
- objective function
- column generation
- decision trees
- dec pomdps
- continuous state
- optimal solution
- planning problems
- partial observability
- stochastic domains
- partially observable domains
- infinite horizon
- markov decision problems
- mobile robot
- supply chain
- sequential decision making problems