The analysis and design of observer-based nonlinear control of depth positioning of a spherical underwater robotic vehicle (URV) is investigated. The observer is required for estimating accurately the unknown state variables in the full-state feedback control laws developed, whereby these control laws can be implemented with the unknown states replaced by their observer estimates. The input-output feedback linearization approach and design techniques are employed. Three approximation schemes for smoothing the signum function in the URV model are developed; these smoothing schemes are required for deriving the linearizing feedback control laws and the related results. Simulation results show that the introduction of observer-based nonlinear control would provide a robust method to stabilize and control the depth position of the URV. © 2011 IEEE.