MotionPlanning inanUncertain Environment: Application toan UnmannedHelicopter

In this workwepresent amethodology for intelligent motion planning inanuncertain environment usinga non-local sensor, suchas a radar.This methodology isapplied toan unmannedhelicopter navigating a cluttered urbanenvironment. We show thattheproblemofmotionplanning ina uncertain environment, undercertain assumptions, canbeposedas theadaptive optimal control ofanuncertain Markov decision process, characterized by a known,control dependent system, andanunknown, control independent environment. Thestrategy formotionplanning then reduces tocomputing thecontrol policy basedonthe current estimate oftheenvironment, alsoknownasthe "certainty equivalence principle" intheadaptive control literature. Ourmethodology allows theinclusion ofa non-local sensor intotheproblemformulation, which significantly accelerates the convergence of the estimation andplanning algorithms. Further weshow thatthemotionplanning andestimation problems, as formulated inthispaperpossess special structure which can be exploited to significantly reducethe computational burdenoftheassociated algorithms. We applythismethodology totheproblemofmotion planning foran unmannedhelicopter ina partially knownmodeloftheTexasA&M campus.