SCS Undergraduate Thesis Topics
|Wennie Tabib||Red Whittaker||Feature Detection on an FPGA|
Fast, accurate, autonomous robot navigation is essential for planetary robots when communication with Earth is delayed, limited or nonexistent. Feature detection is the crux of robot autonomy. It calculates feature descriptors which are the basis of visual odometry and obstacle avoidance. The first step in calculating a motion estimate using visual odometry is to detect features in a pair of stereo camera images mounted on a vehicle. Without feature detection, the robot cannot determine where it is and how to get to its destination. Feature detection is understood, but requires fast computation, which is not available on planetary rovers. The feature slows rover progress. Consequently, the feature is turned off, risk is accepted, and exploration progress is possible. For example, the Mars Exploration Rover was entrapped and lost when feature detection was turned off.
The need is for fast, spaceworthy feature detection. This means near-real-time speed, low power consumption, and radiation tolerance. This research pro- poses to accomplish this vision by implementing feature detection on a Field Programmable Gate Array (FPGA). FPGAs possess the physical attributes for success. The question is whether a feature detection algorithm can be manifested on an FPGA.