Neuromechanical Control of Stability and Maneuverability in Rapidly Running Cockroaches

The extraordinary locomotor capabilities of animals arise from intimately coupled neural and musculo-skeletal systems. Running animals, from cockroaches to humans, all use comparable underlying spring mass dynamics, which simple models suggest provide passive stability without the need for external sensing. In my first chapter, I test if animals can make use of these advantageous dynamics using cockroaches, one of nature's more versatile runners. I show that during rapid locomotion over randomly rough terrain, cockroaches do not alter activation of control muscles via neural feedback. Cockroaches rely instead on a neural feedforward strategy and advantageous mechanics to stabilize in the face of perturbations up to three times their hip height. Still, neural feedback remains critical, particularly for negotiating large perturbations and enabling navigation. This leads to a central challenge in the emerging field of neuromechanics, namely, how mechanics conditions the translation of neural feedback signals to dynamic effects. In the second chapter, I rewrite the neural signals to individual muscles of freely behaving cockroaches. This reveals the control potential specific patterns of neural feedback can have on body dynamics and how these effects depend on mechanical context. Comparable neural feedback can enable (1) linear, precise control of body rotation and translation under static posture conditions, (2) non-linear dynamic control of vertical acceleration during running or (3) rotational control to mediate turning. In the third chapter, I explore mechanisms for this diversity of control potentials by considering the work done by muscles in the context of perturbations and maneuvers. Neural activity in different mechanical contexts results in markedly different and unpredicted muscle functions. Neural feedback acting through a positive mechanical feedback loop can dynamically recruit a single muscle from acting as a brake to behaving as a motor. Finally, in the fourth collaborative chapter, I work with a team of engineers to develop simple control models for antenna-based tactile navigation of complex environments. We consider the combined effects of neural sensing and running dynamics, test these models through neurophysiological recordings of cockroaches' antennae, and instantiate them in a biologically-inspired artificial antenna for robotic navigation.