Publication details

Stochastic Shortest Path with Energy Constraints in POMDPs: (Extended Abstract)

Authors

BRÁZDIL Tomáš NOVOTNÝ Petr CHATTERJEE Krishnendu CHMELÍK Martin GUPTA Anchit

Type Article in Proceedings
Conference Proceedings of the 2016 International Conference on Autonomous Agents & Multiagent Systems
MU Faculty or unit

Faculty of Informatics

Citation
Field Informatics
Keywords POMDP; planning; energy constraints; decision trees
Description We extend the traditional framework of POMDPs to model resource consumption inducing a hard constraint on the behaviour of the model. Resource levels increase and decrease with transitions, and the hard constraint requires that the level remains positive in all steps. We present an algorithm for solving POMDPs with resource levels, developing on existing POMDP solvers. Our second contribution is related to policy representation. For larger POMDPs the policies computed by existing solvers are too large to be understandable, an issue particularly pronounced in POMDPs with resource levels. We present a procedure based on machine learning techniques that extracts important decisions of a policy and outputs its readable representation.
Related projects: