MIT's DARPA Urban Grand Challenge

Back on October 10th, John Leonard gave a Georgia Tech Robotics Institute talk about MIT's DARPA Urban Grand Challenge experience.  The MIT entry, a Land Rover LR3 named Talos, came in fourth place overall (out of 6 finishers and 11 qualifiers).  I thought the most interesting aspect of the design was that it was originally intended to be a "low cost" solution (meaning many $6k SICK lidars, low-cost cameras, and radars), but that ultimately the success of the design hinged on the use of the $75k Velodyne lidar and an equally (or more) expensive Applanix GPS plus Inertial Measurement Unit (IMU) combo.  Regardless, it was an impressive piece of engineering, and they have released much of their code and driving datasets to the public.  Be sure to check out the rest of the post below to see some cool point-cloud visualizations made possible by those phenomenal Velodyne lidars!

The video below shows a visualization of Talos' sensor suite.  The most prominent features are the "sweeping" SICK lidars (the ones that form "lines" in front of the vehicle) and the 360-degree circular swaths from the 64 lasers of the Velodyne. 

 

As John mentioned in the talk, the MIT team's final design hinged on the Velodyne lidar (pictured below on Talos and by itself) to produce the above point-clouds and an Applanix GPS for motion planning and localization.  However, there is now shame in this... virtually all of the top teams used these sensors!

MIT DARPA Urban Grand Challenge  Velodyne Lidar (Laser Range Finder)

If you'd like to learn more about the design of MIT's entry, the perception, and control of the robot, I recommend checking out the impressive Journal of Field Robotics paper, entitled "A Perception-Driven Autonomous Vehicle".  As an aside, MIT's Talos and Cornell's Skynet robots collided during the competition in what has been dubbed "the world's first well-documented collision between two fully-sized autonomous vehicles."  If you'd like to learn more about the details of the first robotic car-crash (probably not the last), check out the Journal of Field Robotics paper, entitled "The MIT-Cornell Collision and Why It Happened".

 

 

 

Comments

Hey Mr. Hizook,

I met the guy at MIT who was in charged of the 18-month Talos project. I was amazed to hear from him that the project cost them (MIT) 4 M, and that they were happy with 4th place. I guess it had to do with being the first time MIT participated in this competition, and not the cost. They are trying to use the Talos' navigation software to control a forklift in a warehouse.

-Carlos

—Anonymous