12 December 2008
Prosilica GC cameras track lanes in experimental robotic vehicle designed by GeorgiaTech University and SAIC.
The Defense Advanced Research Projects Agency (DARPA) Urban Challenge was created in 2004 to promote the development of robotic vehicles following a US Congressional mandate. The latter stipulates that “it shall be a goal of the Armed Forces to achieve the fielding of remotely controlled technology so that by 2015 one third of operational ground combat vehicles are autonomous in order to keep war fighters out of harm’s way”.
The 2007 Urban Challenge held at a former Air Force Base in Victorville, California featured unmanned vehicles operating in a mock 60-mile city environment. Competing vehicles are required to execute simulated military missions including check points, supply missions and obstacles. All participating vehicles navigate through the course aided by various sensors and positioning systems located on the car, and without a driver or the use of a remote control.
The College of Computing at Georgia Tech in Atlanta together with the Science Applications International Corporation (SAIC) entered the race with their fully autonomous vehicle, Sting 1. Sting 1 is a high-end Porsche Cayenne SUV chosen for habitat space, power and its many built-in automation and computer controls. The vehicle was retrofitted for complete computer control of functions such as steering, throttle, brakes, lights and windshield wipers. Upgrades to allow the vehicle to run autonomously included:
The Prosilica GC650C is a fast ultra-compact VGA camera with excellent image quality and sensitivity. Several factors were crucial in its selection: small size, flexibility, reliability and GigE interface for simplified integration and fast data transmission.Both cameras were installed inside a weatherproof housing on either side of the car and facing front. The two GC650C were used to provide lane tracking information for Sting 1. Engineers programmed several models for various types of roads: straight, curvy, and single to four lanes. As it moves, the vehicle analyzes the images provided by the Prosilica cameras and matches the information to a road model to adjust its trajectory in order to drive in the middle of the lane. The cameras were set to image continuously over the duration of the course. The image data was transferred via Cat-5e cable to a single computer for on-board processing and analysis. The software was programmed using Java and C++. Six of the PCs were used to process the information received from the various sensors while the other two acted as servers to send commands to the car.