This project involved the design and fabrication of a self-balancing monopedal robot which is intended to be used as a platform for physically validating simulated risk network based control analysis. A precomputed risk network allows a robot to evaluate the risk that an action will lead to an imminent fall or lead to a state from which the robot will eventually fall after several jumps.' The physical implementation of the simulated robot will allow the theoretical boundaries of safety to be validated. If validated, risk matrix analysis will allow a system to be modeled such that the controller can modify control inputs which would lead falls. The results of physical testing will be used to refine the simulated model. The robot was designed to be as simple as possible while still being capable of operating in three dimensions to study hybrid dynamics and underactuated locomotion. A mechanism with a direct kinematic relation to the output along with a static contact area was designed to allow the ground force profiles to be accurately controlled. In order to utilize the risk network, the force applied by the foot as well as the robot's take-off angle and rate of angular rotation at take-off are key parameters which must be measured and controlled. The robot was be optimized to precisely control these parameters rather than to achieve the longest or highest jump possible as is the objective of other studies.;
展开▼
