Mathematical-based benchmarking and performance improvement for ANT exploration algorithm

Exploration robots are sent inside collapsed building in search of trapped victims. Once inside, they have to explore the entire maze of obstacles and debris as quickly as possible without missing any accessible area by means of the search algorithm. Comparing the performances of emerging search algorithms against known benchmarks, inside realistic simulated fields are crucial for fast victim localisation. In this paper we apply the ridge regression to compute the coefficient of relationship between the ¿ANT victim discovery time step¿ and the ¿search field complexity¿. The estimated coefficients allow us to predict the victim discovery time step of ANT exploration algorithm inside various generated search fields. By means of our coefficients we improve the ANT performances by introducing the extended ANT exploration (EANT). We also evaluate EANT performances by comparing them with the performances of other existing algorithms.