Urban traffic health assessment based on multi-state semi-Markov process

Traffic congestion has become one of the most severe urban challenges, which requires terrific health management. The primary task in health management is to conduct health assessments. In this study, a novel approach is introduced for urban traffic health assessment, which captures the multi-state, multidimensional, and dynamic characteristics of urban traffic systems. We define the network health state of each spatiotemporal window using multidimensional health indicators (reliability, vulnerability, and resilience), and employ a semi-Markov process to model transitions between health states as a multi-state system. The proposed health assessment method is applied to road network in Beijing, and results reveal that 1) four distinct health state types are identified, each with unique multidimensional health characteristics; 2) dynamic transition patterns of health states are learned for different road network regions; and 3) a health ranking of road network regions is developed, suggesting possible personalized management strategies for different regions. Our findings demonstrate the applicability of multi-state modeling in urban traffic systems and provide key methodology for health management of urban traffic.