科研成果详情

Vehicle routing is a challenging problem in smart transport networks. Most of the existing algorithms (if not all) are not always suitable for dynamic traffic scenarios because they do not fully consider traffic congestion. Congestion causes delay, which degrades the performance of routing. We propose a statistical approach to address the issue based on traffic speeds. First, a series of traffic speeds is decomposed into regular and irregular components. The aim is to reduce the unpredictable variance in the components. The first component is estimated using the Holt-Winters algorithm; the estimation of the second component is a challenging part. We introduce a new method to measure congestion probability from the congestion point. This congestion probability is then taken into consideration during the path planning for each trip. The objective is to minimize the travel time and improve the trip's predictability. The effectiveness of the proposed algorithms is evaluated using simulations. The results have shown that using the Holt-Winters algorithm, the components of the level and seasonality of traffic speeds are accurately estimated on a real-world benchmark dataset with a mean squared error (MSE) of 5.08. Simulations also showed that the proposed algorithm improves travel delay and predictability.