Uncertain Low Penetration Rate -- A Practical Issue in Mobile Intelligent Transportation Systems

Low penetration rate is one of the essential issues in the mobile phone based traffic state estimation model. This paper proposes an appropriate genetic algorithm (GA) mechanism to optimize the traffic state estimation model even in cases of low penetration rate. This mechanism also reduces the critical penetration rate, thus improves the error-tolerance as well as the scalability of the traffic state estimation system. The paper also investigates the ANN-based prediction model to overcome the weakness of the GA-based traffic state estimation approach when the penetration rate becomes unacceptably low. In addition, the effect of different level related road segments on the prediction effectiveness is thoroughly discussed. Consequently, this study provides practically useful instructions in verifying the data missing rate at different level related road segments to ensure the prediction accuracy. The experimental evaluations reveal the effectiveness and the robustness of the proposed solutions.