This paper describes a neuro-fuzzy controller for autonomous underwater vehicles (AUVs) of which the dynamics are highly nonlinear, coupled, and time-varying. The neuro-fuzzy controller is based on the fuzzy membership function-based neural networks (FMFNNs) with advantages of fuzzy logics and neural networks, such as inference capability and adoption of human operators' experience with fuzzy logics, and universal approximation and learning capability with neural networks. Unlike other conventional control approaches, the presented FMFNN controller does not require any information about the system, off-line learning procedures, or human intervention to adjust parameters. On-line learning of the FMFNN controller is achieved by using an inner-loop learning scheme and simplified derivatives of the vehicle system. Simulation results show effectiveness of the FMFNN controller for AUVs.