Miniature animal computer interfaces: Applied to studies of insect flight and primate motor pathways.

摘要

The actions of even the most advanced robotic machines are primitive in comparison to mundane tasks performed by animals. Although, we can crudely emulate the mechanics of skeletal joints and muscles with the latest fabrication techniques and materials, we do not yet fully understand the control algorithms innate to biological neuromuscular systems. Stimulus-response experimentation, a traditional tool used by neurobiologists to understand the working principles of the biological systems, is conventionally done with unnaturally constrained or reduced animal preparations with artificial stimuli. Though such experiments have contributed to a wealth of information about animal brains, results obtained under restricted conditions do not reflect the full repertoire of neuromuscular activity that occurs during natural behaviors. For example, electrophysiological experiments to study flapping flight of insects use insects glued on sticks so that desktop instrumentation can record neuromuscular activity via implanted microwires. Similarly, studies of primate motor cortex function have conventionally relied on monkeys performing repetitive, over-trained tasks and confined within a restricted workspace.; To enable experiments in freely behaving organisms in order to understand the neural correlates of intelligent animal behavior, I have developed a family of miniature implantable circuits ("neurochips"). The circuits include: (1) ultra-light data recorders and stimulators for free-flight experiments to reverse engineer flapping insect flight, and (2) autonomous brain-computer-interfaces for primates to understand the functioning of neuromuscular control. My neurobiologist colleagues and I have used the neurochips for a variety of experiments to study in-flight maneuvering in hawkmoths, cataloging brain-motor activity and inducing cortical plasticity in freely behaving macaques. This dissertation describes the neurochip designs and compiles several results from these experiments. The present neurochips have very few recording and stimulation channels, limiting our experimental paradigms. So, I propose future neurochip architectures to scale the number of channels, as well as an analog spike classifier module to discriminate multiple spikes from multiple neural signals.