Experimental Analysis of Different Pheromone Structures in Ant Colony Optimization for Robotic Skin Design

The optimization of the wire routing in an artificial skin for robots consists in selecting a subset of links between adjacent tactile sensors in order to connect them to a finite set of microcontrollers with a minimum cost. The problem has been modelled as a minimum cost Constrained Spanning Forest problem with solution-dependent costs on arcs. The problem is NP-hard. A MIP formulation, which applicability is limited, is given. An Ant Colony Optimization (ACO) algorithm, recently introduced, is also described for tackling large scale problems. This paper introduces several different alternative pheromone structures, whose effectiveness is evaluated through experimental tests, performed on both real and synthetically generated instances.