The CAltAlt and POND Planners

CAltAlt - Conformant AltAlt is a regression planner that searches in belief space with conditional actions to find conformant plans.  The innovation of CAltAlt is its heuristic guidance approach based on planning graphs.  It relies on a Labelled Uncertainty Graph (LUG), a condensed version of the multiple graphs used in CGP, for extracting heuristics to guiding search.  The LUG can make use of a wide variety of mutex relations; the most expressive are similar to CGP. [Contact us for Linux Binary]

POND - Partially Observable Non-Deterministic planner is a progression search belief space planner that uses LAO* search over belief states represented with BDDs.  POND's search is guided by heuristics from the LUG to find conformant and contingent plans. It can find plans that take the form of directed acyclic graphs.  Although POND allows observations and it uses the LUG for heuristics, it does not incorporate observations or knowledge into the heuristics, yet!  Unlike CAltAlt, POND does not use mutexes in its construction of the LUG.  [Linux Binary available]


Papers
Copyright 2005. Last updated: 7/11/05, by Dan Bryce (dan dot bryce at asu dot edu)

Background: Lynx Lake, Prescott, AZ 9/6/03 (a "pond" in Dan's hometown, still looking for a picture of a CAltAlt...)