Performance Evaluation of Undivided Streets: A Clustering-based Approach

抄録

Traffic characteristics and driving culture are the factors that vary across the nations. For instance, the heterogeneity in traffic composition and the loosely-enforced lane discipline make the traffic conditions of India unique and complex as well. Therefore, the adoption of criteria recommended in Highway Capacity Manual of developed nations leads to erroneous results in performance evaluation of Indian urbanstreets. Further, the performance evaluation of ‘undivided’ streets is more challenging as traffic movements of opposite directions tend to impede each other. In this regard, the present study identifies ‘Percent Speed Reduction’ (PSR) asan alternative performance measure and proposes a methodology to assess Level of Service (LOS) based on the same. A sum of twelve hours’ traffic data was collected using videography at mid-block section on undivided street in New Delhi, India. Free Flow Speed (FFS) of the traffic stream was measured employing different techniques however did not show any significant difference in the measurements. K-mean clustering technique was used to classify the observed PSR data into a number of sub-groups. Silhouette method revealed that the cluster members are fitted within the range in the best possible way when the number of cluster is six. Hence, six LOS thresholds corresponding to discrete ranges of PSR were proposed.