Date of Award

2013

Degree Type

Thesis

Degree Name

M.Sc.

Department

Computer Science

First Advisor

Scott D. Goodwin

Keywords

Artificial intelligence, Computer science

Rights

CC BY-NC-ND 4.0

Abstract

In this thesis, we present a new algorithm: Demand Sensitive Map Abstraction (DSMA). DSMA is a special kind of hierarchical pathfinding algorithm in which we vary the granularity of abstraction of the high-level map based on pathfinding request demand associated with various regions in the high level map and the search time of the last path request. Additionally, the low level A* search is not restricted by the boundaries of the high level sectors. By dynamically varying the abstraction we are able to maintain a balance between path quality and search time. We compare DSMA with two variations where the granularity of abstraction is constant; one of those contains maximum granularity throughout (Dense HA*) and the other contains the minimum (Sparse HA*). Our experimental results show that DSMA's performance is a balance between Dense HA* and Sparse HA*. Depending on the resources available DSMA can behave either as Dense HA* or as Sparse HA* or lie somewhere in between. Moreover we do not pre-cache paths at any level, which gives us the added benefit of working with a flexible abstract map without the necessity of changing the pre-cached paths if the low level map changes.

Share

COinS