Sensor based mapping and navigation for underwater robots

This thesis presents a mapping and navigation system intended for an unmanned untethered underwater vehicle. The system utilizes range data obtained from a time of flight sonar operating at 307KHz. The range data, along with an angle measurement of the transducer, is used to generate a simple object map (detected object and its position in two dimensions). The raw range data is filtered using an edge detection algorithm. The edge detection algorithm extracts possible corners from the acoustic data of the scanned environment. The output of the edge detection algorithm is sent to a confidence program. The confidence program determines which of the possible "corners", determined by the edge detection algorithm, are "actual" corners. The output of the confidence program is then used to produce the object map. This object map may be used as the input to an annotated map-builder. The output of the confidence program is then input to the navigation system. The navigation system determines the position of the vehicle relative to a detected object without any a-priori information, which may be used as an input to a path planner and an obstacle avoidance system. The experiments were carried out in a 25 x 30 foot pool and the experimental data processed on a Sun Workstation using Matlab and C generated code for post-processing of the raw acoustical data.