Merging human living spaces with robot workspaces sparks the development of novel control, estimation and perception algorithms with theoretically provable performance and safety guarantees. However, to date, rigorous experimental protocols and equipment for validating the theoretically predicted performance of the proposed algorithms does not exist. A major reason for this is the difficulty associated with reproducing—in a systematic fashion—the dynamic conditions encountered by robots operating in typical human-centric and natural environments. The ultimate goal of this project is to promote human-robot co-existence by developing a unique new fully integrated sensing and control instrument, which will enable researchers to rigorously evaluate the performance of robotic algorithms in response to systematically induced dynamic perturbations that emulate real-world conditions. This award enables the initiation of the effort to realize the envisioned instrument. In this project, a component of the instrument will be realized to support research on legged locomotion using a bipedal robot integrated into a controllable environment realized by a novel instrumented variable impedance treadmill capable of emulating a wide variety of terrain realities. The unique configuration of this component will allow researchers to evaluate robot locomotion control and planning methodologies and will be used to generate sufficiently rich and expressive datasets, which will be widely shared through the internet to assist research in training models of locomotion through machine learning algorithms. The proposed effort will pave the way toward the development of similar instruments for experimentally testing and verifying the strengths and limitations of robotic algorithms.
This award will result in the development of a stand-alone component of a unique shared-use research instrument, which will make possible—for the first time—measurements that will contribute to the creation of novel experimental protocols for the rigorous assessment of robotic locomotion algorithms. This project will involve assessment of the robustness of locomotion controllers to terrain variability and uncertainty by performing synchronous measurements of the (i) motion/kinematics of a novel bipedal robot walking on a unique variable stiffness treadmill, (ii) the interaction forces between the robot’s feet and the treadmill, and (iii) the deflection of the treadmill belt when a compliant environment is simulated. Data of this kind are invaluable to studying and assessing the performance of locomotion control strategies on a wide variety of terrain realities. Perturbations like sudden changes in the stiffness of the terrain underneath a leg—including the case where a leg encounters terrain disruption in the form of step-down or step-up disturbances—can be systematically induced and their effect can be measured by the developed instrument. Such capabilities have never been available in the context of human-sized robotic co-workers.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
