Planning and reasoning for autonomous vehicle control

Abstract

A reasoning system to support the planning and control requirements of an autonomous land vehicle is described. This system is designed specifically to handle diverse terrain with maximal speed, efficacy, and versatility. The hierarchical architecture for this system is presented along with the detailed algorithms, heuristics, and planning methodologies for the component modules. The architecture is structured such that lower-level modules perform tasks requiring greatest immediacy, while higher-level modules perform tasks involving greater assimilation of sensor data, making use of large amounts of a priori knowledge. In describing the component modules of this system, specific techniques for mission planning, map-based route planning, local terrain navigation, and reflexive vehicle control are presented.These techniques have been demonstrated both in a detailed realtime simulation and on a small indoor robotic vehicle.

Global minimum for active contour models: a minimal path approach

“…The consistency condition guarantees that the solution converges to the true one as the grid is refined. This is known not to be the case in general graph search algorithms that suffer from digitization bias due to the metrication error when implemented on a grid (Mitchell et al, 1987;Kiryati and Székely, 1993). This gives a clear advantage to our method over minimal path estimation using graph search.…”

An Efficient Snell's Law Method for Optimal-Path Planning across Multiple Two-Dimensional, Irregular, Homogeneous-Cost Regions

“…The standard algorithm for finding optimal paths on terrain with a known cost map is a dynamicprogramming "wavefront propagation" (Chavez and Meystel 1984;Graglia and Meystel 1987;Parodi 1985;Witkowski 1983;Mitchell et al 1987). (Note that path-finding methods that search a visibility graph of region vertices (Lozano-Perez and Wesley 1979) apply to terrain with only obstacles, not regions of varying cost; and methods that designate "good" areas of free space (Brooks 1983;Crowley 1984;Rueb and Wong 1987) do not give optimal paths.)…”

“…Sixth, although this is rarely recognized, wavefront-propagation algorithms are incapable even in principle of obtaining optimal solutions in the limit when the number of grid cells modeling some fixed terrain area approaches infinity, because of the "digital bias" effect (Mitchell et al 1987), that solution paths not following the "natural" directions of the grid tend to be longer than necessary because of the "zigzagging". To illustrate, consider a starting point inside a forest and a goal point in a field adjacent the forest.…”

“…Certainly it helps to use the A* search algorithm rather than branch-and-bound at a local level (Graglia and Meystel 1987;Mitchell et al 1987), but the time improvement was rather small in a variety of experiments we conducted (Richbourg, 1987). (Note that use of A* does not affect the propagation-direction problem, as the bias arises from inaccurate cost calculation along a path, not in estimating its future potential.)…”

“…(Note that use of A* does not affect the propagation-direction problem, as the bias arises from inaccurate cost calculation along a path, not in estimating its future potential.) A more significant variant (Graglia and Meystel 1987;Mitchell et al 1987) is to allow more complex forms of connectivity between the cells to enable a larger number of possible directions of traversal at the local level. But all these other forms of connectivity require http://faculty.nps.edu/ncrowe/snell2.htm an effectively lower resolution in the terrain representation.…”