Acquisition of ground behavior on the locust model under the virtual physical environment

Abstract

We have studied about ground behavior on the locust model under the virtual physical environment. In the previous study, the locust model acquires jumping behavior by use of neuro-evolution. However, a real locust often uses the jumping and walking behavior more flexibility. In this study, we realize the jumping and walking behavior for the same locust model. Both of those behavior are trained by use of neuro-evolution which is composed of artificial neural network(ANN) and real-coded genetic algorithm(RCGA). We analyze the obtained jumping and walking behavior and show the possibility to acquire flexible behavior. © 2013 Springer-Verlag.