Optimal Search, Location and Tracking of Surface Maritime Targets by a Constellation of Surveillance Satellites

摘要

The key to integrating surveillance operations and architectures is the allocation and tasking of multiple surveillance assets for the purpose of collectively satisfying specified surveillance information requirements subject to time, space, and resource constraints. A generic framework (GAMBIT) for addressing this issue was developed by P.E. Berry & D.A.B. Fogg (DSTO Research Report in draft 2003). The framework assumes that the supporting technologies in terms of communications bandwidth and processing power are in place. The issues associated with the maximal exploitation of space-based surveillance assets are unique due to the nature of the platforms, their sensors, and the communications links to the ground station for target information processing and sensor tasking. This report provides optimal solutions to the problem of dynamically determining the allocation and control of space-based surveillance resources for the purpose of detecting, locating, and tracking surface maritime targets using the GAMBIT formalism. The sensors' performances were modelled in terms of their probabilities of detection and false alarm. Three types of target motion model were considered: deterministic, Gauss-Markov, and conservative. Multiple targets were allowed. The solution progressed by cyclical target motion prediction and Bayesian update. Significant computational efficiency was achieved by use of a localized formulation of the problem. To illustrate the implementation of this process, the use of SAR satellites with several modes allowing different swaths and resolutions was considered. Optimal selection of swaths for target tracking, by choosing the swath for each access that minimizes the expected entropy (MOE) across the region of interest, was demonstrated. Three other MOE were used to select the swath: random, maximum probability, and maximum sum of probability. All were shown to be better than fixed swath or random swath selection. (3 tables, 12 figures, 10 refs.).