Direct Optimization of Robot Parameterization for Trajectory Performance
Much effort within the field of robotics has been made to study and mimic the agility of biological flight. Emulation of bat flight is particularly difficult, as bats utilize numerous independent means of control of both their inertial and aerodynamic characteristics to complete a variety of complex maneuvers. In this thesis, we investigate the viability of enabling a reduced-DoF bat robot to synthesize one such maneuver, inverted perching, by simultaneously and directly optimizing both the configuration and the trajectory of the robot. We begin with a minimal model of a flapping flight system. Noting that longitudinal inertial dynamics represent the dominant behavior for the perch- ing of biological bats, we introduce a single additional degree of actuation: a mass that may be shifted along the longitudinal axis of our system. We use the Lagrangian method to derive the equations of motion for our model, and then construct an augmented system where design parameters, namely link- age masses, are decision variables that are constrained to a constant value. We then reduce our optimization problem to an instance of the direct collocation trajectory optimization method, and find the minimum-time perching robot and trajectory. Our final configuration is able to complete the perching maneuver on a similar timescale to biological bats, suggesting viability of the reduced-DoF configuration.