Using a Genetic Algorithm to Optimise Maritime Surveillance Performed by Space-based Sensors
David M. Lingard
Last modified: August 5, 2011
This study focussed on how to optimise maritime surveillance from space through the use of a genetic algorithm. Small constellations of satellites in Low Earth Orbits (LEO) were considered, hosting sensors such as Synthetic Aperture Radar (SAR). Specific vessel paths were considered and a variety of vessel speeds were analysed.
A genetic algorithm was employed to optimise how a sensor is steered for each pass of the host satellite past a specific vessel path being considered. The fitness measure was the probability of detection (Pd) of the vessels. It was beyond the scope of the study to consider detailed sensor-related issues such as signal-to-clutter ratio and signal-to-noise ratio, and thus it was effectively assumed that a vessel was certainly detected if it was intercepted within a sensor footprint.
A tool was developed to perform the analysis. STK was employed to simulate the constellation and trajectories of the sensor footprints on the Earth’s surface. STK was controlled through an interface built using Java. Information about when the sensors accessed the vessel paths was passed to Matlab where the Pd analysis was conducted, including reformatting and transforming the data, and running the genetic algorithm. Parallel processing was used to run the genetic algorithm more efficiently.
After optimisation of the sensor steering, the Pd was estimated as a function of the vessel speed for a given vessel path. The Measure of Effectiveness (MoE) was the average Pd over a representative range of vessel speeds, such as 4 to 20 knots. It was then possible to estimate the number of satellites required to achieve a satisfactory value of the MoE, for example greater than 95%.
Beyond the scope of this study were other important issues such as launch of the constellation into orbit, cost, data processing, and information handling.