UC San Diego

Many successes in modern robotics can be fundamentally attributed to the standardized tool sets that exist for actuation and sensing with large-scale robots, with the design and control techniques surrounding electromechanical motors serving as a prime example. However, no such tool sets have become standard at the millimeter- and micrometer-scale, leaving actuation and sensing as nontrivial when developing small-scale robotic systems. In this thesis, I develop manufacturing and algorithmic tools for a versatile class of small-scale rotary actuator – rotary pouch motors. I present a design approach that allows for integration of both actuation and sensing into low-profile, laminate linkages of virtually arbitrary length and complexity and demonstrate this approach with the design of a three degree-of-freedom (DOF) spatial manipulator. I also propose a dynamic model of rotary pouch motors and integrate this model into estimation and control laws that are used for real-time joint angle control of a 1-DOF joint. The presented state-estimation algorithm exhibits a root-mean-square (RMS) error of approximately 0.87 degrees and the presented control law demonstrates RMS tracking error of approximately 6.70 degrees when tracking sinusoids up to 2 Hz. Finally, I present a hierarchical scheme for simultaneous joint angle and stiffness control and demonstrate its efficacy in tracking sinusoidal trajectories while maintaining a constant stiffness. By developing these manufacturing, modelling, and control tools for rotary pouch motors, they become more broadly applicable to use in myriad small-scale robotic systems, including insect-scale legged robots and millimeter-scale manipulators.