A stable adaptive control method based on neural networks is proposed for mobile manipulators. Adaptive compensation of the approximation error of a neural network based on a priori knowledge on the dynamic form of the system is considered without knowing the system parameters exactly.