Scalable precomputed search trees

Authors:
Manfred Lau;James Kuffner
Affiliations:
Carnegie Mellon University and JST, ERATO, Igarashi Design Interface Project, Tokyo, Japan;Carnegie Mellon University
Venue:
MIG'10 Proceedings of the Third international conference on Motion in games
Year:
2010

Citing 5
Cited 2

Artificial Intelligence: A Modern Approach

Artificial Intelligence: A Modern Approach
Behavior planning for character animation

Proceedings of the 2005 ACM SIGGRAPH/Eurographics symposium on Computer animation
Planning Algorithms

Planning Algorithms
Precomputed search trees: planning for interactive goal-driven animation

Proceedings of the 2006 ACM SIGGRAPH/Eurographics symposium on Computer animation
Survivability: measuring and ensuring path diversity

ICRA'09 Proceedings of the 2009 IEEE international conference on Robotics and Automation

Precomputed motion maps for unstructured motion capture

EUROSCA'12 Proceedings of the 11th ACM SIGGRAPH / Eurographics conference on Computer Animation
Precomputed motion maps for unstructured motion capture

Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation

Quantified Score

Hi-index	0.00

Visualization

Abstract

The traditional A*-search method builds a search tree of potential solution paths during runtime. An alternative approach is to compute this search tree in advance, and then use it during runtime to efficiently find a solution. Recent work has shown the potential for this idea of precomputation. However, these previous methods do not scale to the memory and time needed for precomputing trees of a reasonable size. The focus of this paper is to take a given set of actions from a navigation scenario, and precompute a search tree that can scale to large planning problems. We show that this precomputation approach can be used to efficiently generate the motions for virtual human-like characters navigating in large environments such as those in games and films. We precompute a search tree incrementally and use a density metric to scatter the paths of the tree evenly among the region we want to build the tree in. We experimentally compare our algorithm with some recent methods for building trees with diversified paths. We also compare our method with traditional A*-search approaches. Our main advantage is a significantly faster runtime, and we show and describe the tradeoffs that we make to achieve this runtime speedup.