Decision making process of hexapods in a model of complex terrains

Abstract

The paper describes behavior of a cognitive control system model, which enables a hexapod to walk in an obstacle-free terrain as well as in a complex terrain including obstacles. This cognitive system model is based on reinforcement learning and assumes the concept of static-stable walking. The decision making process was tested using three different types of terrain models. The results of decision making process trigger actions in the form of changes in the state of six-legged body to maintain stable walking forward. New method have been developed to describe a group of obstacles of different sizes in a complex terrain. The results suggest a relationship between the predefined number of actions and the maximum total walked distance in terrain. In case of the terrain without obstacles, the optimized actions are the same. Thus, the way of moving the trunk and legs in the terrain is always the same and cyclic. The results also indicate that the maximum total walked distance is reduced due to a growing number of obstacles to overcome. The maximum total walked distance is reduced more significantly in the case of overcoming a greater number of small obstacles compared the case of smaller number of large obstacles. The way of moving the trunk and legs in the terrain with large obstacles is acyclic. The methods proposed for the study of the cognitive system and the sensory system of a hexapod, for the simulation of six-legged walking, as well as for the characterization of terrain with obstacles may find application in bioengineering, robotics, military system and other fields.