Optimization of multi-type traffic sensor locations for network-wide link travel time estimation with consideration of their covariance

Due to the propagation of traffic congestion from upstream to downstream links and the uncertainty of path choice behaviours, travel times between different links, particularly adjacent links, are highly correlated in a typical period from day to day. To improve the estimation accuracy of both the mean and covariance of link travel times, a novel measurement is proposed to optimize the locations of multi-type traffic sensors for link travel time estimation. Multi-source data from different types of traffic sensors can be integrated to better estimate link travel time in an entire road network. In practice, the allocation of multi-type traffic sensors is constrained by the total financial budget and should be optimized in accordance with measurement errors and the cost ratio. Numerical examples of synthetic and real road networks are conducted to demonstrate the applications and merits of the proposed multi-type sensor location model with covariance effects.

Automated summary

This study proposes a method to optimize the locations of multi-type traffic sensors in order to improve the accuracy of link travel time estimation. By integrating data from different types of sensors, link travel times in an entire road network can be better estimated. The method takes into account constraints such as total financial budget, measurement errors, and cost ratio.