Mastering Robotic Contact: Combining Mechanisms and Algorithms

Department of Mechanical Engineering & Materials Science Seminar

Dr. Neel Doshi
Harvard University
Intelligence Community

"Mastering Robotic Contact: Combining Mechanisms and Algorithms"

We imagine the robots of tomorrow seamlessly interacting with the world around them, from completing household chores to navigating extraterrestrial surfaces. On the other hand, the robots of today largely avoid unstructured physical contact when possible. Mastering contact will help today's robots bridge this gap, enabling successful interaction with real-world objects and surfaces. Frictional contact, however, poses a major difficulty as it leads to under-actuation, hybrid dynamics, and uncertainty. This results in complex computational problems that must be solved in real-time for robots to reliably interact with the world. In this talk, I will describe how these problems can be efficiently solved through the integration of mechanical and algorithmic intelligence. I will first present my work on millimeter-scale legged robots, such as the Harvard Ambulatory MicroRobot (HAMR). I optimized HAMR's design and developed novel algorithms to plan and control its locomotion, allowing for controllable interaction with the ground. This resulted in more efficient, stable, and fast locomotion, dynamic behaviors like jumping, as well as vertical and inverted climbing. I will then describe how I extended this integrated approach to conventional scale robot manipulation through the design of novel end-effectors (i.e., robot hands) and the development of models and algorithms for dexterous manipulation. I will conclude this talk by discussing my future vision for combining mechanical and algorithmic intelligence to allow robots to reliably interact with the world, enabling intelligent manufacturing processes, autonomous environmental inspection, and on-site assembly and repair.

Bio: Neel Doshi is an Intelligence Community (IC) post-doctoral fellow advised by Professor Alberto Rodriguez at the Massachusetts Institute of Technology, where he focuses on developing methods for planning and controlling contact-rich robotic manipulation and designing novel robot end-effectors. He graduated with a B.S. in Mechanical Engineering (2012) and a M.S. in Robotics (2013) from the University of Pennsylvania. He earned his PhD (2019) from Harvard University under the supervision of Professors Robert Wood and Scott Kuindersma, where he worked on the design of millimeter-scale robots, including the Harvard Ambulatory MicroRobot (HAMR), and created algorithms for planning and controlling the locomotion of legged robots at this scale. Neel's research is broadly focused on developing synergistic mechanical and algorithmic solutions that allow robots to make purposeful and reliable contact with real-world objects and surfaces. He is the recipient of the Intelligence Community post-doctoral fellowship in 2019 and the NDSEG graduate research fellowship in 2013. He is also a co-author on a publication that received the Best Automation Paper award at ICRA 2014, as well as on publications that were nominated for the Best Paper Award at IROS 2017 and ICRA 2018.

Monday, March 15, 2021
Contact Dept. for Zoom details
4:00 - 5:00 pm

Host: Professor Aaron Dollar