Investigating the role of the visual system in solving the traveling salesperson problem

AbstractThis article used an empirical experiment and a computational model to test the hypothesis that humans rely on the visual system to solve the traveling salesperson problem (TSP). We tested two consequences of this hypothesis: (1) humans should perform better on Euclidean TSP than not-Euclidean TSP; (2) a model of the visual system should account for performance in Euclidean TSP. Participants were asked to solve Euclidean or not-Euclidean TSP, and a pyramid model of the visual system was used to solve the same tours as the humans. The results show that deviations from the optimal tour were smaller in Euclidean problems than in not-Euclidean problems, and the fit of the pyramid model to human performance was worse on not-Euclidean problems then on Euclidean problems. These results suggest that participants solve Euclidean problems with the visual system, but that other mechanisms are needed to succesfully solve non--visual problems.


Return to previous page