Optimization of the Memory Weighting Function in Stochastic Functional Self-Organized Sorting Performed by a Team of Autonomous Mobile Agents

The activity of a team of autonomous mobile agents formed by identical “robot-like-ant” individuals capable of performing a random walk through an environment that are able to recognize and move different “objects” is modeled. The emergent desired behavior is a distributed sorting and clustering based only on local information and a memory register that records the past objects encountered. An optimum weighting function for the memory registers is theoretically derived. The optimum time-dependent weighting function allows sorting and clustering of the randomly distributed objects in the shortest time. By maximizing the average speed of a texture feature (the contrast) we check the central assumption, the intermediate steady-states hypothesis, of our theoretical result. It is proved that the algorithm optimization based on maximum speed variation of the contrast feature gives relationships similar to the theoretically derived annealing law.