Improved cooperative planning for air vehicles searching for a ground object
Graham V. Moon · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
This paper considers the problem of controlling a group of air vehicles with imaging sensors in order to search a region for a ground object. This can be formulated as a state-space planning problem where the states represent the possible future location and orientation of the air vehicle and resulting footprint seen by the sensor. Previous approaches have identified the mathematically optimal solution to the problem. However the planning tree within which the optimal solution is found grows very rapidly. For large search areas missions have long duration and it is infeasible, in terms of computation time, to consider all branches of the planning tree. Because of this a common approach is to constrain the solution to solving a sub problem, where plans are only created to a planning horizon. This approach can also be challenging in terms of computation time. This is dependent upon the planning horizon required to give good performance and the amount of co-operation required. Therefore many previous approaches have proposed solving sub-problems of this type sub-optimally. This paper presents an algorithm that utilises a best first search, an optimistic node selection technique, and a novel processing framework. This is then applied to optimally solve a specified sub-problem. The results demonstrate the feasibility of the approach by presenting typical computation times for various planning horizons.