This paper proposes a novel integrated navigation filter based on a combined long baseline/ultra short baseline acoustic positioning system with application to underwater vehicles. With a tightly coupled structure, the position, linear velocity, attitude, and rate gyro bias are estimated, considering the full nonlinear system dynamics without resorting to any algebraic inversion or linearisation techniques. The resulting solution ensures convergence of the estimation error to zero for all initial conditions, exponentially fast. Finally, it is shown, under simulation environment, that the filter achieves very good performance in the presence of sensor noise.