For the second time this summer, Frank Dellaert has been awarded a Test of Time Award. This time, the award comes from Robotics Science and Systems (RSS) and recognizes Dellaert’s papers, Square Root SAM and Square Root SAM: Simultaneous localization and mapping via square root information smoothing. The latter is co-authored by Michael Kaess, an associate research professor at Carnegie Mellon University.
The award recognizes papers published at RSS (or journal versions) that have the highest impact from at least ten years ago or before.
Dellaert was recognized earlier in the summer with the 2020 IEEE ICRA Milestone for his 1998 paper, Monte Carlo Localization for Mobile Robots.
“Receiving a test of time award is amazing and humbling. Being recognized for the RSS paper is special to me because it recognizes work Michael and I have done at Georgia Tech, after I joined as an assistant professor back in 2001. I remember very well giving the talk at the inaugural RSS conference at MIT, in 2005, and even then I felt it was my best work to date,” said Dellaert, a professor in the School of Interactive Computing and Machine Learning Center at Georgia Tech. Dellaert is also affiliated with the Institute for Robotics and Intelligent Machines and the GVU Center.
Published in 2005 and 2006, the papers pioneered an information smoothing approach to the Simultaneous Localization and Mapping (SLAM) problem with square root factorization. They interpreted the problem as a graphical model and as a result created the widely-used GTSAM free software repository.
This work introduced factor graphs as graphical representation for robot mapping, an important application of robotics, such as autonomous driving which is now a multi-billion dollar problem. Factor graphs are also important in aerial vehicles like drones.
When reflecting back on these papers, Dellaert reminds his students that good things take time.
“It took me a long time before I made all the right connections between graphical models and linear algebra, eventually settling on factor graphs as really the most insightful of graphical languages. And it took several years more before we arrived at iSAM, Michael’s Ph.D. work, which incrementally solves the mapping problem,” said Dellaert.
Dellaert and Kaess will deliver a keynote address regarding this work on Tuesday, July 14 at 10:00 am PDT.