A cellular automata ant memory model of foraging in a swarm of robots

This study proposes a cellular automata ant memory model (CAAM) that controls a robot swarm when undertaking the foraging task in a previously known environment with nests. The floor field is well known to all robots, which share the same environment, communicating through the inverted pheromone. This substance is deposited by each swarm robot over every step in searching, which results in a repulsive force between team members. Besides, a short-term memory inspired by Tabu Search is applied to enable robots to remember their last positions and to avoid useless explorations. On the other hand, homing is based on the behavior observed in pedestrian evacuation, resulting in an attractive force through the nests. Moreover, a dynamic information is used to avoid queues of robots and bottlenecks next to the nests. Each robot step is a first choice movement with a stochastic conflict solver, which results in a non-deterministic characteristic to the model. The proposed model was implemented and submitted to several simulations to evaluate its resultant behavior. Different environmental conditions were employed to refine its intrinsic parameters. The results shown that the proposed model is able to perform the foraging task in a competitive way: in searching the swarm perform a good environment coverage and in homing robots are able to find the most adequate nests.