Design and Analysis of a Biped Leg to Survive High-Impact Falls

Abstract

Humans have a remarkable agility and can perform explosive and dynamic motions such as walking, running, jumping and landing. This is in part due to the ability to handle large magnitude ground reaction forces (GRF) in conjunction with outstanding control capabilities of our body. On the other hand robotic bipedal systems have shown limited capabilities. Therefore, performing comparable levels of athleticism and robustness to perform explosive actions as humans, is still a challenge for bipedal systems. In this thesis work we present our efforts to extend the capabilities of bipedal systems to include the ability to perform high impact drop jumps. We define drops jumps as the action starting at a given height above ground, followed by a free-fall phase, and finished by impacting the ground feet first. As a proof of concept we focused on the development of a novel mechanical design of a single robotic leg. Our proposed mechanical design minimizes the overall mass and inertia of the leg, augment actuators' power and implement an impact attenuation system based on a fiberglass leaf spring and a semi-active magneto-rheological damper. A 3D CAD model of the final mechanical design was built, which was further used to develop a comprehensive simulation model of the leg in Matlab's Simscape environment. We draw inspiration from biomechanics and implement an impedance control to perform simulated drop jump experiments. Optimization techniques are used to find optimal gains for the impedance control. Simulation results show the importances and usefulness of having a physical damping mechanism on the leg to perform safe and controllable drop jumps from heights greater than 3 meters. The results of this thesis are intended to be used in the development of a bipedal robot that is capable of dynamic walking, running, and other high impact motions such as drop jumping. In addition to the work on legged robot, we also report preliminary results on the implementation of a nano aerial vehicle testbed for advance control algorithms. The proposed testbed was developed around the open-source Crazyflie Nano Quadcopter by Bitcraze. A motion capture arena is used to track the global position of the quadcopter. The a position trajectory tracking controller was implemented. Trajectory tracking experiments demonstrate the current capability of the proposed control framework to track trajectories within 10cm.