基于浮动车数据的随机路径选择模型参数估计

Parameter Estimation of Stochastic Route Choice Model with Floating Car Data

提出利用浮动车出行数据对城市路网多OD对logit随机路径选择模型参数估计方法.首先采用相对阻抗,实现logit模型参数无量纲化和单一化,将离散多项logit模型转化为相对阻抗的连续概率模型;然后对多OD对的浮动车数据进行分区和权重处理,计算得到概率分布;最后在推导连续型概率密度和累积概率函数基础上,利用最小二乘法进行参数估计.以广州市浮动车数据对方法进行验证,结果拟合优度高,所得参数可有效反映实际路径选择行为.方法简单实用,特别适用于利用海量浮动车数据进行路径选择行为分析.

This study proposes an algorithm to estimate the parameter of Logit route choice model in urban road network with multinomial OD pairs utilizing taxi-floating car data. The relative route impedance is used and the Logit diversion parameter for whole network can be described by one unified dimensionless parameter. The discrete route choice probabilities are converted to continuous form with consideration of the weight of the numbers of trips, and a cumulative function relationship between route choice probability and relative impedance is re-derived from the original Logit route choice model. The least squares method is utilized to estimate of the unified parameter. The numerical example from taxi GPS data of Guangzhou shows that the goodness fit is excellent and the proposed algorithm is effective and convenient to model the route choice behavior with available and massive floating car data.