The Robot Stock Car Autonomous Racing (RoSCAR) project has its inspiration in early stock car racing, where teams competed using automobiles that were unchanged from the original factory configuration. With the playing field leveled, the main discriminant was the ability of the driver. We follow this same model, but of course our RoSCARs drive themselves.
Student teams are provided a common, “stock” hardware platform which they may not modify. It is a completely robot-enabled vehicle, featuring an Asus Xtion sensor for perception, an integrated high resolution optical encoder for velocity estimation, and an onboard 64-bit desktop class computer running Ubuntu Linux. Students are also provided higher-level software interfaces to all of the hardware. By taking hardware differences out of the equation, the main discriminant in a RoSCAR’s performance is the perception, planning, and control algorithms developed and implemented by the teams. This underscores the fact that the greatest challenges faced by today’s robot car developers lie in the software.
Our RoSCAR concept was validated in an experimental course at Lehigh University in the Fall of 2013. Highlights from the final exam (race) are shown in the video below. The winning team demonstrated average lap speeds of up to 5.7 m/s (equivalent to 202 km/h at full scale). We are currently investigating the viability of a regional robotics competition around this concept.