SCS Undergraduate Thesis Topics
|Boris Sofman||Tony Stentz||Obstacle Map construction from Aerial Information for Unmanned Ground Vehicle Navigation|
Unmanned Vehicle navigation has received heavy attention from the robotics community. The ground view of terrain, however, is insufficient for many unmanned ground vehicle purposes. A vehicle navigating at high speeds in off-road environments may be unable to react to negative obstacles such as large holes and cliffs. The PerceptOR project attempts to complement the sensing capabilities of an unmanned ground vehicle with terrain elevation data gathered from an unmanned helicopter. This information can often be misleading. For example, the aerial vehicle will perceive tree canopies as solid, impenetrable obstacles, leading the ground vehicle to entirely avoid areas which it may have been able to navigate with relative ease. A body of water, on the other hand, will be falsely identified as ideal terrain for vehicle navigation.
My research explores techniques to significantly improve automated obstacle detection by utilizing additional sensor information such as color and signal reflectance. I exhibit methods to represent this sensor data in such a way that a neural network can successfully train on a small set of labeled data in order to classify a much larger map. Additionally, I show how these algorithms can be customized for the intended vehicle's capabilities in order to create more accurate obstacle maps that can be then used for path planning.