Evolving and generalising morphologies for locomoting micro-scale robotic agents

Designing locomotive mechanisms for micro-scale robotic systems could enable new approaches to tackling problems such as transporting cargos, or self-assembling into pre-programmed architectures. Morphological factors often play a crucial role in determining the behaviour of micro-systems, yet understanding how to design these aspects optimally is a challenge. This paper explores how the morphology of a multi-cellular micro-robotic agent can be optimised for reliable locomotion using artificial evolution in a stochastic environment. We begin by establishing the theoretical mechanisms that would allow for collective locomotion to emerge from contractile actuations in multiple connected cells.
These principles are used to develop a Cellular Potts model, in order to explore the locomotive performance of morphologies in simulation. Evolved morphologies yield significantly better performance in terms of the reliability of the travel direction and the distance covered, compared to random morphologies. Finally, we demonstrate that patterns in evolved morphologies are robust to small imperfections and generalise well to larger morphologies.