NRI: A Novel Framework for the Hardware and Control Co-design of Dynamic Humanoid Robots with Electric Motors

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

Quadruped robots are now doing real work including exploration of abandoned mines, inspection of power plants, transportation of payloads on rough terrain, and more. The remarkable recent transition from research platforms to useful robots was enabled by fundamental innovations on powerful, inexpensive, and widely available electric motors. But quadrupedal robots cannot use tools created for humans or work in environments designed for the anthropomorphic form, such as narrow corridors or climbing ladders. Instead, capable humanoid robots could help prevent the recurrent injuries and disabilities of firefighters, nurses, or warehouse workers in the US, and save billions of dollars on health care each year. The fundamental challenge preventing the advent of widely adopted humanoid robots is that their mechanical structure and motion control are significantly more complex than those of quadrupedal robots. This project will close this gap by translating the technology that enabled the remarkable success of quadruped robots to the creation of physically capable humanoid robots via a formal and systematic design process. The robot Dash fabricated in this project will realize athletic tasks, such as running and powerlifting, with a performance comparable to humans. The robot will be open-sourced to lower the barrier to entry for research, establish a common platform for benchmarking performance, accelerate the research progress on humanoid robots, and expedite their transition to useful tools in the US. This project will create an integration framework for the co-design of the hardware and control of humanoid robots with electric motors to maximize their physical capabilities. The main fundamental science of this project is a holistic framework for robot integration of dynamic humanoids and at the core of this framework is a novel co-design optimization formulation that is scalable to multiple tasks. The transformative nature of this formulation is amplified by three fundamental scientific contributions of this work: (i) Derivation of performance metrics and design principles tailored to dynamic humanoid robots; (ii) Creation of efficient computational tools for simulation and identification of the complex mechanisms employed by these robots; And (iii) extension of dynamic whole-body control techniques to address self-collision. The open-source humanoid robot Dash will be integrated with this framework for realizing demanding physical tasks including running, powerlifting, and carrying a substantial payload. The lessons learned from this research will be reflected in education through the creation of a course on the design and control of dynamic robots jointly taught by the PIs. 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.