Robust Collective Locomotion with and Without Coordination

I will discuss examples from our recent work on leveraging the power of collectives to function as robust locomoting units. At one extreme, I will discuss our studies of myriapod locomotion; in biological experiments on centipedes, we use advances in neural network trackers (e.g. DeepLabCut) to extract kinematics of limb and body postures vs time. We analyze these dynamics using a geometric formulation of locomotion and show how these animals can optimally coordinate body bending and limb stepping patterns. When instantiated in a robophysical centipede model consisting of repeated two-legged units whose limbs display directional compliance and whose body segments are joined with soft elements, the geometric scheme for coordination of limb and body undulation creates a robot that can traverse diverse complex terrain in open loop. As a counterpoint to tightly coordinated units, I will discuss how groups of worms (worm blobs) can make decisions to avoid environmental threats without presumed system-wide coordination. In particular, the blobs can collectively crawl away from heat sources, which we hypothesize results from an uncoordinated interaction of leading-edge pulling worms and trailing edge lifters, whose individual dynamics are dominated by local heat responses. Implementing analogous control schemes in collections of robots (smarticles) that sense local conditions (light levels) but do not coordinate this information, we demonstrate robust translation without unit-wide coordination. We expect that future robot swarms could move along the (un)coordination axis to flexibly perform complex tasks which might require individual or collective actions.