Abstract
Presented in this thesis is a method for determining locomotion gaits for a modular tetrahedral tensegrity robot in simulation. The biologically inspired tensegrity-based robot is comprised of multiple rigid tetrahedrons connected with a network of adjustable tension members. Locomotion gaits are produced using oscillatory signals from Central Pattern Generators (CPGs) that are optimized by an Evolutionary Algorithm (EA). After initial generation of an EA population of random gaits, as defined by parameters of the CPG, the EA operates on the population in an attempt to find an optimum gait, as defined by the performance, or fitness, of the evolved locomotion. The results show the EA improved the fitness of populations by an average of 32 percent over 10 trials, and up to 51 percent. This is an initial approach to robotic tensegrity locomotion gait production that could be applied to more complex structures, other tensegrity structures or other non-traditional robots.