登入選單

返回Google圖書搜尋

Control of a Biomimetic Robot Lobster with a Synthetic Nervous System

Daniel Herman Blustein

出版	Northeastern University, 2014
URL	http://books.google.com.hk/books?id=LrVkrgEACAAJ&hl=&source=gbs_api

註釋

The biorobotic approach uses robots to simulate biology. This dissertation presents RoboLobster, a biomimetic underwater robot controlled by an electronic nervous system simulation. Using this robot we test neuron network hypotheses related to sensory processing, sensory fusion and walking pattern generation. With a design based on the body structure, biomechanics, and control architecture of the American Lobster, Homarus americanus, RoboLobster serves as a closed-loop biorobotic simulation of neuron network operation that can be run in unpredictable real-world environments. As a secondary contribution, we further investigate arthropod neuroethology using a robotic bee controlled by a computational nervous system simulation. RoboLobster features limb biomechanics similar to those of the model organism. Movement of the robot's eight legs is accomplished by an excitation-contraction coupling scheme in which simulated neuron spikes activate shape memory alloy actuators: muscle analogues. Neuromorphic sensors mimic their biological counterparts and code information in the language of neurons: action potentials. The sensors onboard RoboLobster are optical flow sensors, antennal bend sensors, claw bump sensors, an accelerometer, and a gyroscopic-compass. The inputs from these sensors drive RoboLobster's electronic nervous system which is comprised of mathematically modeled neurons and synapses arranged in network configurations based on the known neural circuitry of the lobster. The electronic nervous system incorporates a variety of simulated neurophysiological components including central pattern generators, command neurons, coordinating synapses, and layered exteroceptive reflex networks. The platform is used to demonstrate plausible implementations of a variety of neurobiological hypotheses, including: lobster walking central pattern generation, arthropod optical flow reflexes, the role of decussating neural connections driving behavior, corollary discharge, and several proposed mechanisms underlying multimodal sensory fusion. RoboLobster extends biological efforts to test neuroscience hypotheses and advances robotic engineering efforts to develop alternative control architectures. This comprehensive biomimetic platform opens the door to a wide range of computational neuroscience experiments and serves as a vehicle capable of operating in the unpredictable conditions of littoral marine environments.