This paper investigates the effect of the penetration rate on the effectiveness of the mobile phone-based traffic state estimation. As a result, the acceptable penetration rate is identified. This recognition is useful for the investigating to bring the traffic state estimation using mobile phones as traffic probes into the real world application. In addition, an adaptive velocity-density estimation model, namely the velocity-density inference circuit, is proposed to improve the accuracy of the average velocity and the density estimations in cases of low penetration rate. Furthermore, a neural network-based prediction model is introduced to assure the effectiveness of the velocity/density estimation when the penetration rate degrades to zero. The experimental evaluations reveal the effectiveness as well as the robustness of the proposed solutions.