Path Planning for Sensing Multiple Targets from an Aircraft
BRIGHAM YOUNG UNIV PROVO UT DEPT OF MECHANICAL ENGINEERING
Pagination or Media Count:
To generate an assignment of tasks that best utilizes a teams resources, it is necessary to know the costs incurred by a team member for doing those tasks. In a cooperative search and destroy scenario, tasks generally require that the vehicles sensor pass over specific known target points, which, to produce the associated costs, requires calculating the path the vehicle will take to sense the various targets. When the targets are far apart, the path-planning problem is trivial. For targets that are closely spaced, however, the problem is much more difficult, and thus is needed the ability to plan paths that sense multiple, closely-spaced targets. Traditional path-planning methods are not well suited for generating paths that sense multiple, closely-spaced targets. Traditional methods focus on connecting some starting point and ending point with a feasible, minimum length path segment. Because an end point must be specified, these methods require too much information about how the path should accomplish its objectives, and hence the complexity of the associated problem is too great for real-time path-planning applications. This thesis introduces the discrete-step path tree, and several methods for finding paths from the tree that accomplish the desired objectives, as solutions to the multiple-target sensing problem. Two of these methods use potential fields to guide the movement of the vehicle through the path tree. However, these methods are problematic and do not produce very good paths. Augmenting the potential-field methods by randomly branching to different parts of the path tree improves the path-length performance, but still not to completely satisfactory levels. The final two methods are based on the path-length performance.
- Numerical Mathematics
- Statistics and Probability
- Target Direction, Range and Position Finding