Search Reduction through Conservative Abstract-Space Based Heuristic - Robotics Institute Carnegie Mellon University

Search Reduction through Conservative Abstract-Space Based Heuristic

Ishani Chatterjee, Maxim Likhachev, and Manuela Veloso
Conference Paper, Proceedings of 10th Annual Symposium on Combinatorial Search (SoCS '17), pp. 161 - 162, June, 2017

Abstract

The efficiency of heuristic search depends dramatically on the quality of the heuristic function. For an optimal heuristic search, heuristics that estimate cost-to-goal better typically lead to faster searches. For a sub-optimal heuristic search such as weighted A*, the search speed depends more on the correlation between the heuristic and the true cost-to-goal. In this extended abstract, we discuss our preliminary work on computing heuristic functions that exploit this fact. In particular, we introduce a many-to-one mapping from an original search space to a conservative abstract space. Edges in the abstract space capture reachability among all corresponding nodes in the original space. We compute a heuristic in the conservative abstract space which when used by the search in the original space reduces the number of searched nodes. Our preliminary results on 3D navigation show that in more complex scenarios the speedup can be dramatic.

BibTeX

@conference{Chatterjee-2017-122730,
author = {Ishani Chatterjee and Maxim Likhachev and Manuela Veloso},
title = {Search Reduction through Conservative Abstract-Space Based Heuristic},
booktitle = {Proceedings of 10th Annual Symposium on Combinatorial Search (SoCS '17)},
year = {2017},
month = {June},
pages = {161 - 162},
}