Design and implementation of cooperative labyrinth discovery algorithms in multi-agent environment

This research focuses on design and implementation of cooperative labyrinth discovery algorithms, specifically, discovering an unexplored maze with multiple robots working collaboratively. Solving a known maze using a single robot is straightforward. The robot looks around and memorizes the structure of the maze and it generates the solution track as a stack of consequent positions from the starting cell to the destination cell. The labyrinth discovery is known as a method of solving the maze when the wall structure is not known. There are various labyrinth discovery algorithms already implemented for a single agent but their extension to cooperating multiple agents is not straight forward and may not produce optimal solutions either. We designed and implemented single agent algorithms, namely Flood Fill (FF) and Modified Flood Fill (MFF), for multi-agent environment. In addition, a cooperative labyrinth discovery algorithm has been implemented based on the ALCK_EF semantic logic. Then we compared their efficiency in theory and practice against the ideal case where agents are aware of the full maze structure. The theoretical comparison is done based on examining the time and space complexity. On the other hand, the experimental comparison examines the total cost for each of those algorithms to solve and discover the maze.