| Biomemetic control of robots has been an increasing topic of research over the past decade. This approach allows life-like robust movements by mimicking the control mechanisms observed in nature. Electronic neurons and synapses can be mapped together to make motor programs for walking, swimming, and flying. This thesis focuses on the application of the Hindmarsh-Rose (HR) neural model to replicate biologically meaningful movement in a novel, bio-hybrid micro-robot. The model's mathematical background is studied and three separate implementations are produced. First, an analog computer method is investigated. Issues solved for this method include scaling for below one volt operation and scaling the model in time for accurate output. A set of tools and a verification board were produced to facilitate a CMOS design. The second method was an implementation of digital neurons on low cost, low pin-count microprocessors. Magnitude scaling of the digital map-based neurons was one technique created for easier processing using only integer variables. An implementation was programmed and proved on a physical demonstration board using 16-bit fixed point variables. The final method was a 65 nm CMOS implementation using an oscillator and an integrator circuit to produce a lower order approximation of the HR model. Benefits include considerably fewer transistors and resistors compared to a previous CMOS circuit. All three serve as a foundation for future research into the area of simple central pattern generators for micro-robots. |
