Modeling shortest path selection of the ant Linepithema humile using psychophysical theory and realistic parameter values

Abstract

The emergence of self-organizing behavior in ants has been modeled in various theoretical approaches in the past decades. One model explains experimental observations in which Argentine ants (Linepithema humile) selected the shorter of two alternative paths from their nest to a food source (shortest path experiments). This model serves as an important example for the emergence of collective behavior and self-organization in biological systems. In addition, it inspired the development of computer algorithms for optimization problems called ant colony optimization (ACO). In the model, a choice function describing how ants react to different pheromone concentrations is fundamental. However, the parameters of the choice function were not deduced experimentally but freely adapted so that the model fitted the observations of the shortest path experiments. Thus, important knowledge was lacking about crucial model assumptions. A recent study on the Argentine ant provided this information by measuring the response of the ants to varying pheromone concentrations. In said study, the above mentioned choice function was fitted to the experimental data and its parameters were deduced. In addition, a psychometric function was fitted to the data and its parameters deduced. Based on these findings, it is possible to test the shortest path model by applying realistic parameter values. Here we present the results of such tests using Monte Carlo simulations of shortest path experiments with Argentine ants. We compare the choice function and the psychometric function, both with parameter values deduced from the above-mentioned experiments. Our results show that by applying the psychometric function, the shortest path experiments can be explained satisfactorily by the model. The study represents the first example of how psychophysical theory can be used to understand and model collective foraging behavior of ants based on trail pheromones. These findings may be important for other models of pheromone guided ant behavior and might inspire improved ACO algorithms.