Realistic Urban Traffic Simulation with Ride-Hailing Services: A Revisit to Network Kernel Density Estimation
App-based ride-hailing services, such as Uber and Lyft, have become popular thanks to technology advancements including smartphones and 4G/5G network. However, little is known about to what degree their operations impact urban traffic since Transportation Network Companies seldom share their ride data due to business and user privacy reasons. Recently, transportation engineering researchers begin to collect data in some large cities trying to understand the transportation impacts of ride-hailing services, to assist transportation planning and policy making. However, (1) there does not exist a general data collection approach applicable to any city, and (2) the studies were based on historical data and cannot project the future easily even though ride-hailing services are developing quickly. In this paper, we introduce our approach to building a digital twin of the transportation in the medium-sized city of Birmingham, Alabama. This digital twin is a transportation simulation model that incorporates transportation modes such as public transits and ride-hailing services, in addition to private vehicles that constitute the majority of Birmingham’s traffic. With this digital twin, transportation engineers can flexibly analyze the impact of ridehailing services under different scenarios, such as if the number of Uber drivers doubles which could happen in the near future. The digital twin may also enable new opportunities, such as being an environment for learning policies with reinforcement learning. To enable realistic transportation simulation, we propose a novel approach to collect Uber ride data that is easy to carry out in any city. This approach collects app screenshots on Uber rides from Uber drivers, and uses crowdsourcing to postprocess these screenshots to extract detailed ride information. We then fit a spatiotemporal distribution of Uber rides using the collected data via network kernel density estimation (KDE). The existing network KDE method is flawed in that the contribution of data samples are not the same, so we propose a new formulation to overcome this problem. The distribution combined with population statistics from census data enable the generation of realistic Uber rides for agent-based simulation.
After obtaining all the extracted ride trajectories, we map them to the underlying road network obtained from OpenStreetMap, using the FMM library for trajectory map matching. Figure below shows a raw trajectory extracted as a sequence of points plotted with a red polyline, and the map-matched trajectory plotted with a green line. We can see that the map-matched trajectory is along the roads and highways, which is properly recovered from the sparse set of raw trajectory points.
We next present some visualization results to summarize our extracted trajectories. We uniformly sampled 100,000 points from the trajectories after map matching, and plotted a heat map using Python’s Folium library. Here, we sample points since plotting the heat map using all trajectory points is too costly for Folium.
We also grouped rides by origin-destination (OD) zipcode areas (defined by TigerWeb6 ) and created an outbound and an inbound trip D3.js interactive visualization. Figure below shows the outbound trip visualization, where a count-threshold bar is used to filter out edges (origin-zipcode, destination-zipcode, count) with low counts, and the darker blue an area is, the more ride origins it contains. The top figure shows the outbound edges from zipcode area 35233 (UAB) highlighted in red, and the bottom figure is when we raise the count-threshold to the highest level, where we can see that UAB and the airport are two biggest ride hotspots.
Click on the images below to go to the YouTube videos.
Urban Traffic Simulation with Ride-Hailing Services around 6:00 PM
Urban Traffic Simulation with Ride-Hailing Services around 9:00 PM
