Hierarchical Behaviour for Object Shape Recognition Using a Swarm of Robots

Abstract

A hierarchical cognitive architecture for robot exploration and recognition of object shape is presented. This cognitive architecture proposes the combination of multiple robot behaviours based on (1) Evolutionary, (2) Fuzzy Logic and (3) Bayesian approaches. First, the Evolutionary approach allows a swarm of robots to locate and reach an object for exploration. Second, Fuzzy Logic is used to control the exploration of the object shape. Third, the Bayesian approach allows the robot to detect the orientation of the walls of the object being explored. Once the exploration process finishes, the swarm of robots determine whether the object has a rectangular or circular shape. This work is validated in a simulated environment and MATLAB using a swarm of E-puck robots. Overall, the experiments demonstrate that simple robots are capable of performing complex tasks through the combination of simple collective behaviours while learning from the interaction with the environment.