Authors: Daniel Vallejo Rodriguez

Venue: Doctoral Dissertation, Texas A&M University
DOI:
Link to Publication

Abstract:
Given a set of fixed objects (called obstacles) distributed in three-dimensional space, and a moving object (called the robot ), the most basic motion planning problem consists of moving the robot from a given initial or start position to a given final or goal position without colliding with the obstacles. The question of whether such a path exists is called a query. If multiple queries need to be solved, then a popular approach is to construct a graph ( roadmap) of representative feasible paths. Queries are solved by connecting the start and the goal to the roadmap and extracting a path between the two connection points. This approach is very good when many queries must be solved in the same static environment because after the roadmap is constructed, usually during preprocessing, queries can often be answered quickly. However, the cost of constructing the roadmap can be very large if the robot or the environment is very complex. If only one or a very few queries need to be solved, then a more direct approach might lead to faster solutions. In this work we concentrate on this situation which is often referred to as single shot motion planning. We propose an adaptive framework for single shot motion planning, i.e., planning without preprocessing. This framework can be used in any situation, and in particular, is suitable for crowded environments in which the solution path must pass through narrow corridors such as maintainability studies in complex three-dimensional CAD models. Our iterative strategy adaptively selects a planning algorithm whose strengths match the current situation, and then, on-line, switches to a different planner when circumstances change. This requires techniques to evaluate the characteristics of the current query, and a set of planners which are characterized so that we can match the query with the best planner for it. Our experimental results with rigid and articulated robots in complex three-dimensional CAD environments show that our strategy solves queries that none of the planners could solve on their own. 

@phdthesis{Rodriguez-aaffss-2000, 
 author = {Daniel Vallejo Rodriguez}, 
 month = {December}, 
 school = {Department of Computer Science and Engineering, Texas A\&M University}, 
 title = {An Adaptive Framework for Single Shot Motion Planning}, 
 year = {2000} 
}