Reliability of San Francisco-Bay Area Road Network

Road network is an important infrastructure in society. On average, every American drives 13,476 miles a year, spending over 200 hours on traveling. Out of the 200 hours, 38 hours are wasted during congestion, wasting $121 billion on fuel cost a year.1,2 Thus, a well-designed road network, which leads to less congestion and thus have significant economic ramifications. Unlike many other networks, road networks are generally static, yet the travel times across edges change constantly, resulting in dynamic traffic patterns. Our group is interested in how congestion occur in real road networks, where a road network is defined as a network of highways for our investigation. To gain an understanding in this field, we reviewed literature related to road networks, with topics ranging from the terminologies used to theoretical modeling of network robustness to game theoretic traffic patterns. Ultimately, we are interested in how traffic changes when the graph structure is altered by removing critical nodes. To answer this question, we used a Bay Area Road Network dataset, which is discussed in more detail in Section 3. We modeled the road network by having roadways as edges and intersections as nodes. In this network, we first identified the critical nodes using 3 different measures: Betweenness Centrality, Mean Closeness Centrality, and Max Closeness Centrality (defined in Section 4.1). Then we simulated how traffic would flow through the area. Finally, we investigated how route assignments of a given set of traffic are impacted after the removal of critical nodes. From our analysis, we found that travel time increased by around 50% when we removed 400 critical