LOGICS TOWARDS POMDP WITH AND/OR WITHOUT REWARDS |
|
|||||
Reachability anaysis in continuous-time Markov decision processes Logics towards POMDP with and/or without rewards Energy consumption in Ad-Hoc and Sensor Networks Stochastic Interfaces An eclipse environment for Modest (finished) Analysis of an Airbag Control Unit (finished) Title Logics towards POMDP (with and/or without rewards) Content The paper [ZHJ05] introduced the logic POCTL* for the discrete-time HMMs, which can be considered as DTMCs but with hidden states. With POCTL* one can express state-based, path-based and believe-state based properties over discrete-time HMMs. The authors presented the model checking algorithm of POCTL* over discrete-time HMMs.
De Alfaro introduced the model checking algorithm of the logic PCTL* over the model DTMDPs, where a state have to choose first an action nondeterministically, and then a successor state according a probabilistic distribution based on the selected action. This thesis should extend the logics POCTL* for HMMs and PCTL* for DTMDPs such that it can be interpreted over POMDPs, the partially observable Markov decision processes. Also model checking algorithms should be investigated. Status Available.
Contact References
|
||||||
Dependable Systems & Software Group | Department of Computer Science | Universität des Saarlandes |