Multi-robot exploration using dynamic fuzzy cognitive maps and ant colony optimization

Abstract

An application field of Multi-Robot Systems (MRS) is within victim rescue operations. The main challenge faced by disaster rescue teams is response time. The chances of finding survivors decrease significantly over time and dramatically decrease after 48 hours. In this context, the motivation of this work is to present an MRS inspired by the concepts of swarm robotics to rescue victims in unknown environments. In this case, the robots are unaware of the search area boundaries and obstacles, knowing the number of victims to be rescued as a stopping criterion for the simulations made in Matlab®. Therefore, three approaches inheriting the main aspects of fuzzy logic are used based on previous works: a fuzzy logic controller (FLC), a dynamic fuzzy cognitive map (DFCM) controller, and a DFCM inspired by the ant colony optimization metaheuristic (DFCM- ACO). The proposed task simulates real life disaster rescue operations, or even humans lost in unfamiliar environments such as forests. The simulations were performed in three environments in order to test the overall robustness against unpredictable situations, autonomy, explored area and processing time for both approaches using a subsumption-based architecture. In general, the results suggest that the DFCM-based MRS approaches are able to complete the tasks consuming less processing time, with robots travelling shorter distances to explore a similar environment to the FLC approach and with the DFCM-ACO presenting balanced results between the other techniques. Finally, future works are outlined. © 2020 IEEE.