PAO for planning with hidden state

Abstract

Abstract-We describe a heuristic search algorithm for generating optimal plans in a new class of decision problem, characterised by the incorporation of hidden state. The approach exploits the nature of the hidden state to reduce the state space by orders of magnitude. It then interleaves heuristic expansion of the reduced space with forwards and backwards propagation phases to produce a solution in a fraction of the time required by other techniques. Results are provided on an outdoor path planning application.

An AO* Based Exact Algorithm for the Canadian Traveler Problem

“…Another AO * -based algorithm for CTP is the PAO * Algorithm presented in Ferguson et al (2004). PAO * shares certain basic characteristics with CAO * and BAO * because all three are based on the classical AO * search.…”

“…The goal here is to find a policy that decides what and where to disambiguate en route so as to minimize the expected length of the traversal. Several 97 heuristics and approximation algorithms have been introduced for CTP in the literature (Baglietto et al 2003, Xu et al 2009, Eyerich et al 2009) and optimal algorithms for certain special cases of CTP have been proposed (Ferguson et al 2004, Nikolova and Karger 2008, Bnaya et al 2011.…”

Penalty-Based Algorithms for the Stochastic Obstacle Scene Problem

“…In CTP, the goal is to find the minimum expected length path over a finite graph whose edges are marked with their respective probabilities of being traversable and each edge's status can be discovered dynamically when encountered. SOSP and CTP have practical applications in important probabilistic path-planning environments such as robot navigation in stochastic domains (Blei and Kaelbling 1999, Ferguson et al 2004, Likhachev et al 2005, minefield countermeasures (Smith 1995, Witherspoon et al 1995, and adaptive traffic routing (Fawcett andRobinson 2000, Gao andChabini 2006). In fact, both problems as well as closely related ones have gained considerable attention recentlysee, e.g., Nikolova and Karger (2008), Eyerich et al (2009), Likhachev and Stentz (2009), Xu et al (2009), Aksakalli and Ceyhan (2012).…”

Planning with imperfect information

“…To find these cells, coined 'pinch points' [10], we generate a set of cells that could reasonably be encountered by an agent navigating to the goal and look for key members of this set. Figure 2 outlines the process.…”

“…In [10] the algorithm PAO* was introduced, which solves planning problems involving hidden state such as pinch points. PAO* is applied to an adjacency graph containing the robot position, the goal position, and the pinch points in the environment.…”