NRI: Collaborative Research:Integrating Physics Models and Control Methodologies for Enhanced Legged Locomotion on Yielding Terrain

NRI: Collaborative Research: Integrating Physics Models and Control Methodologies for Enhanced Legged Locomotion on Yielding Terrain PI: Paul B. Umbanhowar (NWU); Co-PIs: Daniel I. Goldman (Georgia Tech) and Kevin M. Lynch (NWU) Autonomous robots cruise the skies, sail the seas, and roll along roads with performance superior to most animals. We propose to examine movement on yielding natural substrates like sand, snow and grass where the performance of the best robots lags far behind that exhibited by most animals. Unlike air and water, whose mechanical properties change little across the Earth, the mechanical properties of natural substrates vary spatially in both magnitude and kind: snow can be firm or powdery; and rocks can roll, sink, or remain fixed when tread upon. These issues are often questions of scale: to a camel, sand is a yielding, effectively continuous material, but to an ant it appears as a rugged surface of rolling and slipping rocks. Dramatic spatial variations in mechanical response presents two main challenges. First, how can robotic locomotor strategies adapt to maintain effective movement over the effectively innumerable substrate types? Second, how can path planning be accomplished to move groups of robots effectively when mechanical properties are unknown prior to direct interaction and change upon interaction? To advance the state-of-the art in robotic locomotion on natural substrates, our Science of Cyber-Physical Systems focused proposal will examine control strategies for model locomotors on natural substrates. We will employ a threefold strategy that closely combines analysis and theory, robot experiments and biological studies. Simulations and analysis will be compared to validated substrate simulations and experiments. We will characterize the ability of various controllers (including robust, adaptive and potentially novel bio-inspired controllers) to match target motion templates as well as generate new gaits. Biological studies will examine how animals transition between gaits on different substrates and cope with novel terrain. Substrate information obtained from controllers will characterize ground properties and guide trajectory planning. Science of Cyber-Physical Systems: real-time control and adaptation, locomotion, natural substrates Intellectual Merit Our proposed work has broad applicability to a wide range of robot design and control issues that are important and potentially transformative. It is a significant undertaking, but we have the advantage of significant experience and resources. A deeper understanding of how to move effectively on substrates with different kinds of responses will radically enhance our ability to design and control robots for a wide range of terrain, and our specific model system will provide a new paradigm for hybrid dynamics in locomotion. Broader Impacts Applications: Design and control of robots for search and rescue, planetary and space exploration, highly mobile personal robots and other operations requiring traversal of complex terrains. International outreach: Graduate students will have the opportunity to participate in the Hands-On Research in Complex Systems Advanced Study Institute which conducts 2 week long project based training workshops for educators in developing areas. Undergraduate education: Undergraduates will participate in the research as REU students and in other capacities. Several recent undergraduates working in the labs at Northwestern and Georgia Tech have gone on to PhD study in engineering, robotics, and physic