交通流时间序列的复杂度测量

为了定量分析交通流系统的复杂性,引入算法复杂度和近似熵,通过速度时间序列的算法复杂度估计系统周期性成分的比率,在重构序列时通过取多个划分区间来提高算法复杂度的估计能力.计算近似熵时,先由速度序列得到速度变化率序列以去除趋势,然后通过速度变化率序列的近似熵估计系统在结构变化上的复杂性.对实测交通流数据序列的计算表明:在序列长度超过600时可以得到算法复杂度,序列长超过300时可以得到近似熵;交通流的算法复杂度和近似熵在同步状态时较低,拥挤状态时增大,在自由状态时最大.因此,不同的算法复杂度和近似熵对应不同状态下的交通流,算法复杂度能分析较长的交通流序列,近似熵可以分析较短的交通流序列.

Complexity measure of traffic flow time series

In order to quantitatively analyze traffic flow complexity, the Kolmogorov complexity(Kc) and the approximate entropy(ApEn) were adopted, the system periodic ingredient ratio was estimated by using the Kc of speed time series, and the Kc evaluation ability was improved by dividing reconstructing time series into more sectors. In the ApEn computation, the time series of speed change rate was got to remove the tendency of speed time series, then the complexity of system structure change was estimated by using...