Shape memory alloys have become the most promising construction material for smart actuators in diverse applications. However, their highly nonlinear hysteresis behaviour sets the main barriers in precise motion control. This paper presents an adaptive MKP control scheme based on a feed-forward compensator using the inverse of the modified KP model and the selective parameter adaptation law, to realize dynamic hysteresis compensation for real-time applications. The performance of the proposed control scheme was evaluated experimentally. The results indicate that such an improved adaptive control scheme is characterized by the reduced computation, rapid parameter convergence and effective hysteresis linearization.