Travel time variability and congestion assessment for origin–destination clusters through the experience of mobility companies

Reliability is a measure of network performance that reflects the ability of the network to provide predictable travel times. Deviations from planned travel times can increase travel costs for users. To improve the system’s performance, it is crucial to identify sources of unreliability, particularly the location on the network of unreliable performance. The large amount of travel time data recorded by Transportation Network Providers (TNPs) in recent years has enabled researchers to study the performance of entire networks. In this study, a real-world dataset provided by TNPs in Chicago is used to determine time of day, and day of week distribution of travel time per unit distance for origin–destination (OD) pairs. Eight measures of reliability are calculated for OD pairs in the network. Standard deviation (SD), planning time index (PTI), and on-time measure (PR) are used for a network-wide comparison of reliability performance. K-means clustering is performed on more than 21.3 million trips to divide 3,450 eligible OD pairs in the Chicago network into three groups with low, medium, and high intensity of each reliability metric. Lastly, metrics in each cluster of SD, PTI, and PR are compared. The results show that ranking PTI and PR is not sufficient for identifying unreliable/congested OD pairs in the network. Approaches for comparing reliability performance over different periods of the day for the same segment and over different segments in the network are discussed, along with network-wide measures of reliability.