Multi-objective optimization for placing airspace surveillance observers

Abstract

Reconnaissance is an important aspect of military planning. Tools that help an alysts monitor and make informed choices are vital for avoiding costly situations. The use of ground-based radar sensors is a common method for monitoring for both land-based and airborne threats. Manually finding optimal locations to install sen sors within an area of terrain can be difficult and time intensive, particularly when multiple objectives exist. The purpose of this thesis is to implement and compare two heuristic algorithms for automatically generating a set of optimal locations for airspace surveillance sensors. The algorithms seek to find solutions that maximize both total area coverage and coverage of a specific area of interest. They also seek so lutions that minimize sensor overlap and price. The research problem was formulated into a multi-objective optimization. The two algorithms tested include the NSGA-II and a multi-objective Ant Colony Algorithm (MOACO). A population-halving aug mentation and the Multi-resolution Approach (MRA) developed by Heyns [1] were also applied to see if algorithm run time could be reduced without impacting final solution quality. The NSGA-II outperformed the MOACO algorithm with respect to diversity of the final solution set, however the algorithms performed similarly with respect to run time and convergence. It was found that population-halving and the MRA could result in computation time reduction for the tested scenario, however not at a significant level.