Lab Reopening in the Time of Covid19: Cassie Resurgence
Our lab closed on March 20, 2020 under the State of Michigan's “Stay Home, Stay Safe" order. Thousands of labs across the University of Michigan campus closed over a 72 hour period. It was stressful, and as you can imagine, there was a bit of miscommunication. For a 24 hour period, it seemed that our labs would be "sanitized" during our absence. Since we had no idea what that meant, we decided that Cassie Blue needed to "Stay Home, Stay Safe" as well. We loaded up a very expensive robot and took her off campus. On May 26, we were allowed to re-open our laboratory. After thoroughly cleaning the lab, disinfecting tools and surfaces, developing and getting approval for new safe operation procedures, we then re-organized our work areas to respect social distancing requirements and brought Cassie back to the laboratory.
During the roughly two months we were working remotely, the lab's members got a lot done. Papers were written, dissertation proposals were composed, and plans for a new course, ROB 101, Computational Linear Algebra, were developed with colleagues. In addition, one of us (Yukai Gong) found the lockdown to his liking! He needed the long period of quiet to work through some new ideas for how to control 3D bipedal robots. He got his mathematical ideas sharpened enough to be coded up and tested them thoroughly on inverted pendulum models, a 5-link model of Rabbit, and eventually, Cassie Blue. The first week of the lab's opening was spent on "hacking" Cassie's battery and power system so that the robot could operate on a pair of LiPo hobbyist batteries. The sad truth is that early on in the Covid epidemic, the factory in Asia where Cassie's batteries are built was shut down. Our last battery failed the same day that Agility Robotics warned us of the factory's fate.
Once we had Cassie powered up, we could work on execution time of the control code. We had that done by Monday of the following week. Yukai spent the next several days testing Cassie with an older controller and his new controller, discovering a few coding errors, and taking his ideas across the simulation-reality chasm. Remarkably, this was a very quick process with his new controller. By Thursday, Cassie's performance was impressively better than we had achieved before. On Friday and Saturday, we moved outdoors and shot an hour of really good video, of which a few minutes are given here. We'll post more next week.
Nest steps: This video represents preliminary testing. We expect better results in the near future and then we'll write it all up.
We want to give a big shout-out to Associate Dean of Research Steven Ceccio for his tireless work to safely close down the huge research operation that is represented by the College of Engineering at the University of Michigan, and the even bigger lift to re-open it under circumstances that none of us imagined when we were ringing in the New Year.