Reactive obstacle avoidance (OA) is an important step in attaining greater autonomy in autonomous underwater vehicles (AUV). For AUVs that conduct underwater surveys, avoidance of uncharted obstacles can improve vehicle survivability. This paper discusses initial experiments at the Center for AUV Research in obstacle detection and avoidance using the Naval Postgraduate School ARIES AUV with the Blueview Blazed Array forward looking sonar. It includes a discussion on evaluating OA optimality, autopilot control design and sonar image processing. It concludes with a description of successful results from a recent demonstration