Telescope scheduling

Abstract

This dissertation involves heuristics implemented to tackle the Multiple Telescope Scheduling problem. It includes scheduling of proposals submitted by astrophysicists whilst taking into consideration constraints such as telescope lag, visibility and priority of a celestial object. The program generates a schedule that will try to maximize the utilization of the telescopes. The problem is tackled by comparing the results of two heuristics. One heuristic implemented is the genetic algorithm whilst the other heuristic is the memetic algorithm. The memetic algorithm is similar to the genetic algorithm however uses a separate local search to improve the fitness at each generation. The local search algorithm implemented is a steepest ascent hill climbing algorithm which uses two different types of neighbourhood search techniques. One technique tries to improve the order of the celestial objects whilst the other tries to find other targets that fit within the schedule. The functions implemented include two types of selection, two types mutation, two types' crossover and two types' hill climbing techniques. The results were generated using randomly generated dummy data and each algorithm was run with all possible permutations of the available functions. The results demonstrate that the memetic algorithm improves the convergence time. It also improves the priority constraint however the genetic algorithm gave the best schedule by visiting one celestial object more than the memetic algorithm.