The integration of inaccurate remotely operated vehicle (ROV) model data obtained by computational fluid dynamics for control is presented. Since the ROV is highly nonlinear and uncertain, a sliding-mode control (SMC) system using a direction-based genetic algorithm (GA) and fuzzy inference mechanism is proposed. The GA influences the right evolutionary step and direction of the SMC parameters subjected to uncertainties in the evolutionary process. The effectiveness of reducing the sensitivity of the proposed control scheme to model parameters and external disturbance is verified by simulations and sea trial. The results demonstrate that the proposed controller performed better with less chattering in position responses than SMC without GA-fuzzy optimization, fuzzy logic controller, and proportional-integral derivative.