Obstacle Avoidance Approaches for Autonomous Underwater Vehicle: Simulation and Experimental Results

Autonomous underwater vehicles (AUVs) operate in unknown underwater environments and must be able to avoid submerged obstacles such as cliffs, wrecks, and seabed changes. This paper proposes a methodology for obstacle avoidance by AUVs that are equipped with forward-looking sonars (FLSs). The data collected from two FLSs placed in horizontal and vertical orientation are processed in real time to provide obstacle detection information in the xz- and xy-planes, respectively. Due to the necessity of maintaining constant height when employing sidescan sonar (SSS) and lower energy consumption, horizontal avoidance maneuvers are preferred over vertical ones. For the horizontal obstacle avoidance, a two-layer obstacle avoidance algorithm (OAA) contains two different processes, an OAA based on Bandler and Kohout (BK) product of fuzzy relation used as a preplanning method and a reactive approach based on potential field and edge detection methods. The preplanning technique has a clear advantage on a line segment of the path, and the reactive method is more efficient on a curved segment. In case a horizontal approach cannot find a path to safely avoid the obstacle, the reactive vertical approach is activated. A seabed gradient detection technique that allows prediction of seabed and altitude changes up to 40 m ahead of the AUV is presented. Simulation and experimental results clearly demonstrate that the proposed methodology enables AUVs to navigate safely through obstacles and provide crucial information about the seafloor terrain changes.