|
|
Coyote Robotics Enters the DARPA 2007 Urban Challenge Coyote Robotics aims to compete in the 2007 Urban Challenge by utilizing well-established vision and probabilistic robotics techniques to build a well designed vehicle that is capable of completing the 60 mile urban course in under 6 hours. Coyote Robotics, LLC has focused on vision and robotics research for over three years with over 20 years of combined software development expertise. Coyote Robotics partnered with two teams in the 2005 Grand Challenge, one of which was selected to compete in the NQE and completed 2 miles of the NQE test track using solely Coyote Robotics vision software for obstacle detection. Building on the foundation laid over previous years work, the team’s approach to the Urban Challenge 2007 will center around the use of camera sensors with common vision techniques to sense the vehicle’s surroundings, probabilistic techniques such as SLAM for mapping, and probabilistic and discrete search techniques for planning. The techniques include the stereo vision based mapping and obstacle detection developed for the previous Grand Challenge, established probabilistic techniques for lane following and moving obstacle tracking, and extensions to probabilistic SLAM techniques to consolidate all of these inputs. Our approach focuses on the use of cameras over LADAR to leverage the rich image data and other advantages provided by cameras. Since the Grand Challenge 2005, we have continued our development efforts using a smaller test platform. This smaller test platform is a 100 lb. electric ATV named Bravo. We have been testing our software on this vehicle at an area of several unused acres in a park in the Seattle area. All obstacles in the area are artificial placed by us to simulate different urban situations. With this vehicle, it is easy for us to test our software more readily at smaller locations.
|
Copyright (c) 2004-2006 Coyote Robotics, LLC
