作者: Shlomo Zilberstein , Alan Carlin
关键词:
摘要: Anytime algorithms allow a system to trade solution quality for computation time. In previous work, monitoring techniques have been developed agents stop the at "right" time so as optimize given time-dependent utility function. However, these results apply only single-agent case. this paper we analyze problems that arise when several solve components of larger problem, each using an anytime algorithm. Monitoring in case is more challenging agent uncertain about progress made far by others. We develop formal framework decentralized monitoring, establish complexity interesting variants and propose one. Finally, show can be applied flow planning problems.