Using a genetic algorithm for multitarget tracking

The author has devised a technique that uses a genetic algorithm (GA) to address the multiscan assignment problem in multitarget tracking. The assignment problem involves taking detections from one or more sensors over multiple time intervals and dividing them into groups or tracks. A unique set of assignments (a hypothesis) can be expressed as a list which allows the space of possible hypotheses to be searched using a GA. In this paper, he describes how a GA can maintain a population of hypotheses and how it continuously updates this population as more scans of data arrive. Simulation results and comparisons with other track-assignment methods are presented