Bingyu Zhao

Bingyu Zhao

Researcher 
Office: 445 Davis Hall
Email: bz247@berkeley.edu

 

My name is Bingyu Zhao 赵冰玉 and I am currently a postdoc in the Soga group. I work on city-scale traffic simulations and use them as a tool to answer engineering-related questions, such as:

  • How much carbon emissions can we save by paving the roads with eco-friendly materials or implementing eco-friendly pavement maintenance strategy?
  • What are the interactions and interdependencies of multiple infrastructure systems in disaster scenarios (e.g., earthquake and wildfire) and how the failure of one component affects the evacuation and relocation of the residents?

My interest in multi-infrastructure system began with pavements. I was analyzing the pavement degradation data (potholes, cracks, roughness) for city-scale roads in London and San Francisco during my PhD and the analysis of the SF pavement degradation data using spatially regression is documented in this paper. My postdoc is partly founded by NIPPO Corporation, the leader in pavement construction and material production in Japan, through which opportunity I am involved in building a pavement degradation model for major roads in Tokyo.

One key hurdle I encountered in explaining the difference in pavement degradation rates across road segments is to include correct pavement usage history (i.e., vehicle loading). Due to the lack of such data, simulations of city-scale traffic are resorted to in order to obtain the typical vehicle volumes per road . This work is founded on the previous static city-scale simulation models in our group by Gerry Casey and makes use of the efficient queue-based Dijkstra’s path finding algorithm implementation by Krishna Kumar. By incorporating traffic simulations into city-scale pavement degradation and usage stage emission modeling, impacts can be evaluated in regards to strategies such as eco-routing and eco-friendly pavement maintenance scheduling in reducing CO2 emissions of the road transport system.

As a further improvement to the static/quasi-dynamic traffic assignment model in capturing more transient traffic phenomena, I am currently working on developing dynamic mesoscopic simulations for the San Francisco Bay Area. Dynamic models are especially suitable for studying disaster evacuation events with highly variable traffic conditions, while the mesoscopic approach has the unique advantage of combining individual decision making process with real-world macroscopic observations. Examples of its applications include identifying evacuation bottlenecks in earthquakes and wildfires, which are realistic threats in California and other places in the world. Our projects related to evacuation include studying the interdependencies of the water pipelines and highway bridges in earthquake scenarios as well as the role of the communication infrastructures in wildfire evacuations.

Pavement condition change in San Francisco (based on San Francisco Public Works data)
Street-level pavement degradation rate. (a)-(c): Pavement age as the explanatory factor. (d)-(f): Cumulative traffic load as the explanatory factor. (a), (d): regression by pavement material type/functional class categories. (b), (e): regression by individual street. (c), (f): regression constrained by spatial proximity of the streets.
Traffic volumes in 15 minute time slices for San Francisco, based on quasi-dynamic traffic simulations.