Leveraging design for collective phototaxis in morphological swarm robotics

Morphological computing-the use of a robot's physical design to facilitate task execution-has emerged as a promising strategy in swarm robotics. Here, through experiments and simulations, we show that its success critically depends on both the robots' control policy and the details of their physical design. We study a swarm of Kilobots embedded in an exoskeleton whose geometry controls the propensity of robots to align or anti-align with the external force they experience. Experimentally, the contrast between the two morphologies in the success rate of a simple phototactic task becomes dramatic when stopping is forbidden and robots can only slow down. Using a faithful physical model of the robots' self-aligning dynamics, numerical simulations reveal that efficient phototaxis requires fine tuning of the self-alignment strength around a sweet spot. More broadly, varying this parameter unlocks a rich repertoire of collective behaviors.