Dynamic neural field as framework for behaviour coordination in mobile robots

Abstract

Behaviour based navigation frameworks present the need of mechanisms for behaviour coordination. Algorithms inspired by processing principles of the human brain can provide effective solutions to the coordination problem. Here we present the use of the dynamic neural field, a recurrent neural network, as behaviour coordination layer in robotics navigation algorithms. We test the control loop with the humanoid robot NAO in different contexts. Results show that the use of the dynamic neural field allows the generation of context-dependent effective behaviours.