The paper proposes a robust online adaptive neural network control scheme for an automated treadmill system. The proposed control scheme is based on Feedback-Error Learning Approach (FELA), by using which the plant Jacobian calculation problem is avoided. Modification of the learning algorithm is proposed to solve the overtraining issue, guaranteeing to system stability and system convergence. As an adaptive neural network controller can adapt itself to deal with system uncertainties and external disturbances, this scheme is very suitable for treadmill exercise regulation when the model of the exerciser is unknown or inaccurate. In this study, exercise intensity (measured by heart rate) is regulated by simultaneously manipulating both treadmill speed and gradient in order to achieve fast tracking for which a single input multi output (SIMO) adaptive neural network controller has been designed. Real-time experiment result confirms that robust performance for nonlinear multivariable system under model uncertainties and unknown external disturbances can indeed be achieved.