This paper introduces StarlETH, a compliant quadrupedal robot that is designed to study fast, efficient, and versatile locomotion. The platform is fully actuated with high compliant series elastic actuation, making the system torque controllable and at the same time well suited for highly dynamic maneuvers. We additionally emphasize key elements of a powerful real time control and simulation environment. The work is concluded with a number of experiments that demonstrate the performance of the presented hardware and controllers.

This paper presents the actuation system of the robotic leg ScarlETH. It was developed specifically for a quadrupedal robot and is designed to achieve fast position control as well as accurate joint torque control. It introduces strong passive dynamics to create an efficient running behavior. High spring compliance with low damping in combination with a cascaded, motor velocity based, control structure was successfully tested in simulation and experiments. Final tests with the entire leg demonstrate that the system can perform a hopping motion providing only positive actuator power.

Legged robots rely on an accurate calibration of the system's kinematics for reliable motion tracking of dynamic gaits and for precise foot placement when moving in rough terrain. In our automatic foot-eye calibration approach, a monocular camera is attached to the robot and observes the robot's moving feet, which are equipped with Augmented Reality (AR) markers. The measurements are used to formulate a non-linear least squares problem over a fixed time window in order to estimate the 33 unknown parameters. This is efficiently solved with the Levenberg-Marquardt algorithm and we get estimates for both the kinematic and the camera parameters. The approach is successfully evaluated on a real quadruped robot.

To enhance the well known, self-stabilizing effects of Spring Loaded Inverted Pendulum (SLIP) models, researchers have proposed a variety of dead-beat controllers that adjust model parameters (angel of attack, spring stiffness), such that a disturbance is rejected within a single step. While such laws can be nicely encoded for disturbances in hopping height (by using the time of flight as a measure of vertical position), they suffer from substantial drift due to the missing information about the forward velocity and inaccuracies in the actual system model. Without requiring additional complex sensors, we propose a method to estimate the forward velocity of a SLIP model based solely on measuring the time of stance. This method is additionally able to perform realtime parameter estimation, which paves the road to implement a full state dead-beat controller that can reject arbitrary disturbances even in the presence of model and sensor errors.

This paper investigates the mechanical benefits of employing a passive foot segment to improve energetic efficiency in legged running. The proposed lightweight design significantly reduces impact and damping losses, while simultaneously allowing for a natural-looking stance configuration. Actuator input and ankle spring properties were optimized in simulation and successfully tested in 2D running experiments.