There has been some nice media coverage of our Tensegrity Robotics research “recently”. I’m a bit behind on updating this blog, so “recent” really means “in the last 6 months” or so.
Most exciting is an interview by Zygote Quarterly. This is a magazine dedicated to bio-inspired design, and the interview gives some nice insights into how I’ve found inspiration in fusing engineering and biological inspiration in developing the field of tensegrity robotics. And since it is a design magazine, its really pretty with some nice inspirational photos. Worth the read!
Next, back in January the BBC released a nice short video about our research. I always appreciate when someone does independent research and pulls together a story including prior videos and information we have released. It is so refreshing to see given how much of online media is slapped together with minimal effort and is often wildly inaccurate, especially around science topics. As usual, the BBC continues to maintain standards for real content. Thanks!
And, now to give this post some more fun technical content — back in May we presented a Tensegrity Deep Learning paper at the International Conference on Robotics and Automation (ICRA) that was developed jointing with our colleagues at UC Berkeley who led the algorithm development. In it we demonstrated the first example of learned continuous locomotion on the actual SUPERball robot. What is particularly amazing is that we were able to do all the training in simulation, and have the learned policy work directly on the real robot. This is unusual in the world of robotics, and I believe that it highlight the value of flexible, compliant robots like our tensegrity robot — the compliance that makes it resilient to unexpected contact also makes it robust to approximately tuned controllers. You can read the paper, find the code, and more on the project website.
0 Responses
Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.