UC Berkeley

Maneuverability in animals is unparalleled when compared to the most maneuverable human-engineered mobile robot. Maneuverability arises in part from animals' ability to integrate multimodal sensory information with an ongoing motor program while interacting within a spatiotemporally complex world. Complicating this integration, actions from the nervous system must operate through the mechanics of the body. Since sensors and muscles are fused to a mechanical frame, mechanical processing occurs at both at the sensory (input) and motor (output) levels. To reveal the basic organization of the neural and mechanical parts of organisms during locomotion, I studied high-speed sensorimotor tasks in a remarkably maneuverable insect, the cockroach, which integrates sensory information to navigate through irregular, unpredictable environments.

Animals can expend energy to acquire information by emitting signals or moving sensory structures. However, it is not clear if the energy from locomotion, itself, could permit a different form of sensing, in which animals transfer energy from movement to reconfigure a passive sensor. In the first chapter, I demonstrate that cockroaches can transfer the self-generated energy from locomotion to actively control the state of the antenna via passive mechanical elements, with important effects on body control. This chapter advances our current understanding of sensorimotor integration during rapid running by showing how the whole body, not just the sensor, can participate in sensory acquisition.

Information flow from individual sensory units operating on locomotion-driven appendages to the generation of motor patterns is not well understood. The nervous system must rapidly integrate sensory information from noisy channels while constrained by neural conduction delays. When executing high-speed wall following using their antennae, cockroaches presumably integrate information between self and obstacles to generate appropriate turns, preventing collisions. Previous work on modeling high-speed wall following within a control theoretic framework predicted that a sensory controller for antenna tactile sensing of wall position (P) and the derivative of position (D) was sufficient for control of the body. I hypothesized that individual mechanoreceptive units along the antenna were tuned to enable stable running. Extracellular multi-unit recordings revealed P and D sensitivity and variable-latency responses, suggesting the antenna may function as a delay line. In the second chapter, I show how individual sensor units distributed on the antenna precondition neural signals for the control of high-speed turning.

Since sensors of animals are embedded within the body, they must function through the mechanics of the body. In Chapter 3, I studied mechanical properties of the primary tactile sensors of cockroaches, the antennae, using experimental and engineering approaches. I revealed how both the static and dynamic properties of the antenna may influence sensory acquisition during quasi-static and dynamic sensorimotor tasks. Further elucidation of antennal mechanical tuning will lead to new hypotheses, integrating distributed mechanosensory inputs from a dynamic sensory appendage operating on a moving body.

During rapid escape from predators, the neuromechanical system of animals is pushed to operate closer to its limits. When operating at such extremes, small animals are true escape artists benefiting from enhanced maneuverability, in part due to scaling. In Chapter 4, I show a novel neuromechanical strategy used by the cockroach P. americana and the gecko H. platyrus which may facilitate their escape when encountering a gap. Both species ran rapidly at 12-15 body lengths-per-second toward a ledge without braking, dove off the ledge, attached their feet by claws like a grappling hook, and used a pendulum-like motion that can exceed one meter-per-second to swing around to an inverted position under the ledge, out of sight. In cockroaches, I show that the behavior is mediated by a rapid claw-engagement reflex initiated during the fall. Finally, I show how the novel behavior has inspired the design of a small, hexapedal robot that can assist rescuers during natural and human-made disasters.