Design and Analysis of Proximate Mechanisms for Cooperative Transport in Real Robots

This paper describes a set of experiments in which a homogeneous group of real e-puck robots is required to coordinate their actions in order to transport cuboid objects that are too heavy to be moved by single robots. The agents controllers are dynamic neural networks synthesised through evolutionary computation techniques. To run these experiments, we designed, built, and mounted on the robots a new sensor that returns the agent displacement on the x/y plane. In this object transport scenario, this sensor generates useful feedback on the consequences of the robot actions, helping the robots to perceive whether their pushing forces are aligned with the object movement. The results of our experiments indicated that the best evolved controller can effectively operate on real robots. The group transport strategies turned out to be robust and scalable to effectively operate in a variety of conditions in which we vary physical characteristics of the object and group cardinality. From a biological perspective, the results of this study indicate that the perception of the object movement could explain how natural organisms manage to coordinate their actions to transport heavy items