Having recovered from our exhilirating (and exhausting) trip to the DARPA Robotics Challenge (DRC) Trials in Florida, we’ve been mulling over the results of the Trials and looking back on the simulation-based Virtual Robotics Challenge (VRC) that we held in the summer.
As you’ll recall, the goals of the VRC were two-fold. First, we wanted to identify the teams from around the world who are the best at writing software for remote supervisory control of a humanoid robot in disaster response situations. After the VRC, the top seven teams were given access to an Atlas robot from Boston Dynamics, with which they then competed in the Trials. And because they only had a few months between receiving their robot
and heading to the Trials, the second goal of the VRC was to force teams to build and test their software in simulation so that they could be up and running with a working system as soon as they took delivery of the Atlas hardware.
Looking at the results, we’re feeling pretty good about the VRC. At the Trials, which saw sixteen teams compete, five of the top eight teams were previously top finishers in the VRC (the top eight from the Trials qualified for continued DARPA funding). And their relative performance at the Trials almost matched their relative performance in the VRC:
| Team | DRC Trials finish | VRC finish |
| IHMC Robotics | 2nd | 1st |
| MIT | 4th | 3rd |
| TRACLabs | 6th | 4th |
| WRECS | 6th | 2nd |
| TROOPER | 8th | 8th |
So a majority of the top teams at the Trials had first competed and won in simulation, and they were then able to transition their software from simulation to hardware in a very short period of time (they had also received a very capable Atlas robot, which surely contributed to their successes).
We’re proud of these results and we feel that they bode well for the use of simulation in future robotics projects, in competitions, and in education.
It’s also worth noting that team WRECS started out as an unfunded “Track C” team, yet placed 2nd in the VRC and 6th in the Trials, outperforming most of the fully funded teams along the way. Perhaps we’ll see more of this kind of non-traditional source selection in future programs.