基于隐Markov模型的短时交通崩溃事件预测

Short-term traffic breakdown prediction using a hidden Markov model

交通崩溃事件会造成道路通行能力下降,成为导致城市快速路拥堵的主要原因之一,精准的短时交通崩溃事件预测在交通管理与控制中具有重要意义。该文以美国加州高速公路性能评估系统(PeMS)提供的交通流数据为基础,对道路的崩溃状态进行了分级定义,并以道路崩溃状态为隐状态、道路占有率为显状态,结合二者之间的联系,建立了隐Markov模型。通过数理统计,对模型参数进行了学习,最后采用Viterbi算法对该模型进行了求解,实现了快速路交通崩溃事件的预测。预测结果与实际数据的对比表明:该方法能预测大部分的交通崩溃事件。

Traffic breakdown reduces road capacity as one of the main factors causing congestion on urban expressways. Accurate short-term traffic breakdown predictions on urban expressways are becoming more and more important because of their vital role in traffic management and control. Traffic flow data was obtained from the Caltrans performance measurement system （PeMS） with traffic breakdown states classified by a lane-based method. A Hidden Markov model （HMM） is then established with the traffic breakdown state as the hidden state and the road occupancy as the observed state with the Viterbi algorithm to solve the problem. The traffic breakdowns were successfully predicted to show that the HHM accurately predicts short-term traffic breakdowns.