@MASTERSTHESIS\{IMM2012-06483,
    author       = "A. H. Johnsen",
    title        = "Modeling and Planning for Uncertainty - within Mission Operations",
    year         = "2012",
    school       = "Technical University of Denmark, {DTU} Informatics, {E-}mail: reception@imm.dtu.dk",
    address      = "Asmussens Alle, Building 305, {DK-}2800 Kgs. Lyngby, Denmark",
    type         = "",
    note         = "{DTU} supervisor: Thomas Bolander, tobo@imm.dtu.dk, {DTU} Informatics",
    url          = "http://www.imm.dtu.dk/English.aspx",
    abstract     = "Part of the thesis has been a study of dierent approaches for planning under uncertainty. The primary focus has been the planning approaches, ”Planning Based on Markov Decision Processes (MDP’s), Planning for Extended Goals with Progression of Computation Tree Logic (CTL), and Epistemic Planning.” Transition systems in general has also been investigated.
The basic steps with respect to using the planning approaches have been explained briefly. This includes the general model, goal type, planning problem, planning approach, solution type, and agent architecture.
The result of the investigation of the planning approaches was a comparison of the dierent advantages and disadvantages. The focus was mainly expressibility with respect to uncertainty. The most notable findings were:
• State transitions can not be partially observable during plan execution for any of the approaches (that observers the transitions made).
• The probabilistic transition systems could model likelihood on action outcomes.
It should be noted that the findings are purely based on the approaches as they have been specified in this report. There are many extensions/alterations to the approaches which improve on some of the limitations found in this thesis.
The DTUsat2 satellite has been studied as part of the thesis. The study of the satellite aimed to identify some of the planning problems with respect to DTUsat2. The planning problems that was looked at are problems where dealing with uncertainty is paramount to correct planning and operation within the domain.
Recharging the battery of the DTUsat2 satellite was one of these planning problems. The approaches, ”Planning Based on {MDP}’s, Planning for Extended Goals with Progression of {CTL,} and Epistemic Planning,” was tested with respect to the battery recharging planning problem.
The most notable findings regarding the models and the test executions were:
• It is necessary to model some kind of likelihood on action outcomes to avoid
unrealistic plans
• There are states that needs to be avoided for safety critical reasons. There is a risk of the battery exploding if recharging is done when the temperature of the battery is below 0 C or above 45 C."
}