Comparative study of fitness function in genetic algorithm for optimal site allocation using Lidar

Abstract

An autonomous genetic algorithm has been implemented that uses Geospatial data to locate the most feasible locations for setting up base camps in a disaster scenario such as a flood where evacuation and rescue efforts are of primary importance. Modern day geographical information system packages do not incorporate genetic algorithm capabilities for solving the location allocation problem. In this paper, a genetic algorithm was introduced that uses a domain specific objective function, combining spatial aspects of the data with an evaluation of the feasibility of locations of base camps. Lidar data has been used to exclude the selection of locations that are infeasible. Incorporating Lidar data with the genetic algorithm reduces search space and returns results that correspond to locations at ground level. The implemented genetic algorithm is tested using two different fitness functions. Experiments reveal that using sector selection as fitness function yields better results than maximizing distance. For sector selection as a fitness function, the coverage area increases by 23.5% and overlapping area decreases by 88%