基于射频数据的道路交通流路径识别优化模型

针对非RFID(Radio Frequency Identification)覆盖道路交通流路径识别误差较大等问题,本文提出基于FCD(Floating Car Data)校核下RFID道路断面交通流路径识别优化组合模型。首先,利用平移不变小波变换将RFID初始数据切分为可追溯交通流、非追溯交通流及随机项;然后,根据统计路段中浮动车数量将路段分为Full、Defect、Null这3类,并建立FCD-RFID追溯路径模型识别可追溯交通流路径构成,同时,提出考虑出行时间、道路等级和驾驶偏好因素的综合成本阻抗效用函数,通过路径感知随机用户平衡分配模型估算非追溯交通流与随机项路径;最后,通过路径叠加识别断面交通流最终路径构成。结果表明：相较于单一RFID交通流路径识别,组合模型具有更高精度,MAE(Mean Absolute Error)为 72 辆,较单一 RFID 算法下降 62.5%,MRE(MeanRelative Error)为9.5%,下降72.2%;在非RFID覆盖校核道路中,组合模型MRE为13.3%,较单一RFID算法下降82.0%,有效验证了本文模型的可行性及适用性。

An Optimization Model of Road Traffic Flow PathIdentification Based on RFID Data

To reduce the error of traffic flow path identification on non-RFID (Radio Frequency Identification) coveredroads, an optimization model of traffic flow path identification on RFID roads is proposed based on Floating Car Data(FCD)verification. Firstly, the initial RFID data is categorized into traceable traffic flow, non-traceable traffic flow, andrandom items by TIDWT. According to the number of floating cars in the statistical section, the road is classified intothree categories: Full, Defect, and Null. A FCD-RFID tracing path model is established to identify the composition oftraceable traffic flow paths. At the same time, a comprehensive cost function is proposed that considers travel time,road grade, and driving preference. The non-traceable traffic flow and random items path are estimated by RPL-OSUE. Finally, the final path composition of road section traffic flow is identified by path superposition. The resultsshow that, compared with the single RFID traffic flow path recognition, the combined model has higher accuracy. TheMean Absolute Error (MAE) is 72 vehicles, which is 62.5% lower than the single RFID algorithm, and the MeanRelative Error (MRE) reaches 9.5%, which is 72.2% lower than the single RFID algorithm. In the non-RFID coveredroads, the MRE of the combined model is 13.3% , which is 82.0% lower than that of the single RFID algorithm,showing feasibility and applicability of the model.