矩阵估计真的能还原真实OD?基于模拟实验的实证

Can Matrix Estimation Reveal the Actual OD Matrix Effectively?Empirical Evidence based on Simulation Experiments

围绕OD矩阵估计技术是否能还原真实OD这一问题,设计了基于仿真实验的研究方法进行实证研究.大型网络交通模型和小型网络交通模型的研究结果均显示初始输入矩阵是OD估计技术可用性的关键,且对大型交通模型来说构造初始矩阵法近乎失效;路段流量模拟值与观测流量的相关系数不能保证结果的可靠性;考虑转向交通流量的强约束条件下,小型交通模型OD估计结果在观测流量、GEH值等指标上表现近乎完美,但与真实矩阵之间仍存在一定差异.总之,OD估计技术能发挥一定的实践作用但并不完美,追求高质量的初始OD矩阵是该技术有效应用的关键所在.

A simulation experiment-based research method was designed to empirically investigate the question of whether the OD matriXestimation technique can reveal the real OD.It is found that the initial input matriXis the key of OD estimation technique and the initial matriXmethod is nearly invalid for large traffic models;the correlation coefficient between observed traffic and simulated values of road sections does not guarantee the reliability of the results.Under the condition of considering the strong constraint of turning traffic flow,the OD estimation results of the small traffic model are nearly perfect in the test indexes such as the observed traffic GEH value,but there are still some differences with the real matrix.In conclusion,the OD estimation technique can play a partial role in searching the actual trip distribution but is not perfect,and the pursuit of a high-quality initial OD matriXis the right way to use.