The travel survey method based on cellular positioning technologies has been proven to be an effective approach for developing OD information. However, existing methods are primarily based on rule-based algorithms and the accuracy is estimated by limited studies. Meanwhile the cellular positioning error caused by wireless communication disturbance has not aroused enough attention, which has a significant impact on the OD extraction accuracy. This paper explores a method for extracting OD from cellular dataset using a spatial-temporal density-based clustering algorithm (ST-DBCAN). Then an innovative comprehensive traffic and wireless communication simulation platform is established and optimized with 4G-LTE protocols and channel model. The improved platform can construct more realistic wireless communication disturbance environment, which enables the simulated mobile phone dataset resembles the ground truth. Finally, the accuracy and performance of the new method for extracting OD is evaluated based on the platform. The results show that the accuracy of extracted OD information is affected by two factors: the distance from mobile phone to the centroid of TAZs and location area in the city (urban center or suburb). In urban center, the OD extraction error percentage of trip ends near the centroid of TAZs is 2% to 8% and those at the edge is 90% to 93%. In suburb area, the corresponding ratio is 90% to 96% for trip ends at the edge and 4% to 18% for trip ends near the centroid of TAZs.

CICTP