Autonomous underwater vehicles (AUVs) are used for long range undersea operations such as oceanographic exploration and surveying. They typically resemble a torpedo in that most have a propulsion unit and control surfaces, and must therefore move forwards to manoeuvre. Such vehicles are called flight vehicles. This work describes techniques which are candidates for control of flight AUVs and identifies controllers used on some existing vehicles. It is noted that no one single controller type is predominant, and that it is difficult to compare their various merits as they are tested on different vehicles under different conditions. As a basis for comparison, the low-cost, experimental flight vehicle Sub-zero II was developed. A 1m-long vehicle, it has the same actuator and sensor functionality as a full-sized vehicle and is used as a testbed to study autopilots for AUVs. Although it has a thin data cable which links it to an on-shore computer, it is shown that the presence of the cable has a negligible effect on controller performance and thus controllers developed for untethered vehicles can be tested on Subzero II. Classical linear/PID-based, fuzzy logic, and sliding mode control were selected as they are in use on other existing vehicles. Following development using a comprehensive simulation program, these controllers were tested using the experimental vehicle. The findings from extensive pool tests were: fuzzy logic had potential but that significant work was required in tuning, sliding mode control gave the best performance when detailed and accurate system information was available, and where such information was not available then classical control was the best alternative.