NSF Award
9361990
9361990 Weiman Efficient sensory perception and processing are crucial to mobile robot navigation. As sensor technology advances, conventional processing methods become a bottleneck, crippling real time operation. Transitions Research Corporation has recently develop a low-cost LIDAR system capable of generating 10,000 samples per second, challenging conventional sensor data processing rates. This research project will efficiently and robustly compress sensory range-bearing data by applying the log-Hough transform to recognize walls and characterize them parametrically. Walls are the dominant navigation reference features in built environments. The log-Hough transform converts a long chain of sensor hits on a wall to a single data point representing line parameters. This succinct data can be used for localization against a priori maps without overloading communications channels to the navigator. In Phase I, LIDAR data will be gathered and analyzed with respect to range and bearing resolution; software will implement the log-Hough transform for these data. In Phase II, the method will be implemented in a real time prototype navigation system running on TRC's HelpMate mobile robot. In Phase III, the method will be incorporated into a commercial product for mobile robots navigation and obstacle avoidance.
