Control and study of bio-inspired quadrupedal gaits on an underactuated miniature robot

This paper presents a linear quadratic Gaussian (LQG) controller for controlling the gait of a miniature, foldable quadruped robot with individually actuated and controlled legs (MinIAQ-III). The controller is implemented on a palm-size robot made by folding an acetate sheet. MinIAQ-III has four DC motors for actuation and four rotary sensors for feedback. It is one of the few untethered robots on a miniature scale capable of working with different gaits with the help of its individually-actuated legs and the developed controller. The presented LQG controller controls each leg's positions and rotational speeds by measuring the positions and estimating the rotational speeds, respectively. With the precise gait control on the robot, we demonstrate different gaits inspired by quadrupeds in nature and compare the simulation and experiment results for some of the gaits. An extensive simulation environment developed for robot dynamics helps us to predict the locomotion behavior of the robot in various environments. The match between the simulation and the experiment results shows that the proposed LQG controller can successfully control the miniature robot's gaits. We also conduct a case study that shows the potential to use the simulation to achieve different robot behavior. In a case study, we present our robot performing a prancing similar to horses. We use the simulation environment to find the required motor configuration phases and physical parameters, which can make our robot prance. After finding the parameters in simulation, we replicate the configuration in our robot and observe the robot making the same moves as the simulation.