This paper presents the experimental results obtained in the Unmanned Vehicles Laboratory (LVNT) of University of São Paulo working with the guidance and control of the underactuated Autonomous Underwater Vehicle (AUV) Pirajuba. In former works, in addition to the identification of a mathematical model, H∞ techniques were used to synthesize and simulate a controller for the vehicle, subjected to modelling uncertainties, disturbances and measurement noises. In this work, experimental results are presented to validate the proposed approaches. Moreover, a centralized multivariable controller is investigated, seeking to improve the performance of the vehicle in manoeuvres in six-degree-of-freedom. Results obtained ratify the good performance of the H∞ approach in complex control systems involving important nonlinearities and uncertainties. © IFAC.