Mission-Phasing Techniques for Constrained Agents in Stochastic Environments.

Mission Phasing;Resource Allocation;Constrained Agent;Deliberation Scheduling;Computer Science;Engineering;Computer Science & Engineering

Wu, JianhuiTeneketzis, Demosthenis ;

University of Michigan

关键词: Mission Phasing;    Resource Allocation;    Constrained Agent;    Deliberation Scheduling;    Computer Science;    Engineering;    Computer Science & Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/60650/jianhuiw_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Resource constraints restrict the set of actions that an agent can take, such that the agent mightnot be able to perform all its desired tasks. Computational time limitations restrict the number ofstates that an agent can model and reason over, such that the agent might not be able to formulatea policy that can respond to all possible eventualities. This work argues that, in eithersituation, one effective way of improving the agent;;s performance is to adopt a phasing strategy.Resource-constrained agents can choose to reconfigure resources and switch action sets for handlingupcoming events better when moving from phase to phase; time-limited agents can choose to focuscomputation on high-value phases and to exploit additional computation time during the execution ofearlier phases to improve solutions for future phases.This dissertation consists of two parts, corresponding to the aforementioned resource constraintsand computational time limitations. The first part of the dissertation focuses on the developmentof automated resource-driven mission-phasing techniques for agents operating inresource-constrained environments. We designed a suite of algorithms which not only can findsolutions to optimize the use of predefined phase-switching points, but can also automaticallydetermine where to establish such points, accounting for the cost of creating them, in complexstochastic environments. By formulating the coupled problems of mission decomposition, resourceconfiguration, and policy formulation into a single compact mathematical formulation, the presentedalgorithms can effectively exploit problem structure and often considerably reduce computationalcost for finding exact solutions.The second part of this dissertation is the design of computation-driven mission-phasing techniquesfor time-critical systems. We developed a new deliberation scheduling approach, which cansimultaneously solve the coupled problems of deciding both when to deliberate given its cost, andwhich phase decision procedures to execute during deliberation intervals. Meanwhile, we designed aheuristic search method to effectively utilize the allocated time within each phase. As illustratedin experimental results, the computation-driven mission-phasing techniques, whichextend problem decomposition techniques with the across-phase deliberation scheduling andinner-phase heuristic search methods mentioned above, can help an agent generate a betterpolicy within time limit.