On the Mark Information Monotonicity in Random Disambiguation Paths
Xugang Ye, Carey E. Priebe, Donniell E. Fishkind, Lowell Abrams · 2008
We consider the problem of navigating an agent to safely and swiftly traverse a nondeterministic network along a random disambiguation path (RDP). Other than assuming that the probability of existence of any nondeterministic is known as a priori, we assume that a sensor characterized by two conditional distribution functions is available to generate a maker for each nondeterministic arc after an observation. We study whether the improvement on the sensor quality incurs the reduction of the total cost (traveling cost plus disambiguation cost) the agent pays to reach the target given a navigation protocol and in the same terrain. We introduce the concepts of stochastic ordering of sensors and mark information monotonicity (MIM) of protocols. We propose two types of concrete navigation protocols: threshold protocol and penalty protocol and introduce some preliminary analytical results. We also introduce some Monte Carlo simulation results in minefield application.