CAREER: Computational Design for Robust Legged Robots

点击翻译按钮获取中文摘要

This project aims to democratize robotics by making engineering design intuitive and accessible through computation. Currently robot design is a time and skill intensive process requiring years of training and multiple iterations of prototyping and testing. Unlocking the full potential of ubiquitous robotics requires new intelligent design tools that support experienced and novice users alike in the design process, from ideation to prototyping to validation stages. The resulting methods will have the potential to revolutionize the way robotic devices are created, increasing people's ability to design and use custom devices while reducing the time spent on this customization. New flexibility and availability of robotic systems will improve the integration of these technologies into everyday society. Professional engineers will gain the ability to perform systems-level analyses and verification of their designs, streamlining their work, encouraging more creativity, and ultimately producing more efficient and effective machines. In education, this project will inform next generation scientists and engineers by providing intuitive tools for exploring engineering concepts.The project will synthesize ideas from mechanics, computational geometry, robotics, and data clustering and compression to create a computational framework for designing, fabricating, and controlling legged robots that are robust to task, environment, and fabrication uncertainties. While the application will be legged locomotion, the work will focus on computational frameworks and insights that can be generalized to broader areas of robot design and customization. Unlike static objects, legged robots rely on complex interactions with their surrounding environment to locomote, and their performance is therefore very dependent on their physical designs. Compared with existing computational design tools, which focus on optimizing a single metric, this work will therefore evaluate the robustness of a proposed design and its suitability for a practical locomotion task. The project will produce 1) modular, graph-based representations for a robot's combined geometry, kinematics, and joint motion, 2) new hierarchical data clustering and compression techniques for semantic representation and evaluation of a design, and 3) algorithms for design synthesis that provide explainable designs based on evaluated efficiency-robustness tradeoffs. The integrated education plan also brings together ideas from engineering and computer science to incorporate numerical simulations and design techniques into courses at the undergraduate and graduate level, form an open-source robot design community, and encourage broader participation in STEM through outreach to K-12 students and teachers.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.