This research is motivated by the increasingly diverse mobility modes emerging in urban region and the highly heterogeneous traffic flow associated with it. Even though there exists numerous traffic flow models that characterize traffic evolution of only one vehicle class (homogeneous traffic flow) which takes similar amount of roadway spaces and shares the similar speed range, there are relatively few models that apply to heterogeneous traffic flows in urban areas. Traffic estimation is challenging also because dynamic information of each vehicle class cannot be fully obtained directly using sensor technologies available today. To this end, I am interested in investigating quantitatively to what extend the modeling details provided in heterogeneous model can improve traffic estimation. Precisely, my research goal is to construct a flow-model based estimator by combining streaming data with a heterogeneous flow model to sequentially estimate traffic state when the flow is composed of a mix of passenger vehicles and smaller-sized vehicles (e.g., bikes and motorbikes).
The US Environmental Protection Agency identifies that urban heat islands can negatively impact a community’s environment and quality of life. Using low cost urban sensing networks, it is possible to measure the impacts of mitigation strategies in communities at a fine-grained scale, informing context-aware policies and infrastructure design. However, fine-grained city-scale data analysis is complicated by common, tedious data cleaning tasks such as removing outliers and imputing missing data. To address the challenge of data cleaning, this article introduces a robust low-rank tensor factorization method to automatically correct anomalies and impute missing entries for high-dimensional urban environmental datasets. We validate the method on a synthetically degraded National Oceanic and Atmospheric Administration temperature dataset, with a recovery error of 4%, and apply it to the Array of Things city-scale sensor network in Chicago, IL.
This project explores methods for model parameter estimation on a car-following model using data collected from Adaptive Cruise Control (ACC) enabled vehicles. The proposed methods are batch method least-squares and an online particle filter. Numerical experiments demonstrate the accuracy and computational performance of the methods relative to a commonly used simulation-based optimization approach. The methods are also assessed on empirical data collected from a 2019 model year ACC vehicle driven in a highway environment. Speed, space gap, and relative velocity data are recorded directly from the factory-installed radar unit via the vehicle’s CAN bus. The least-squares method has the fastest run-time performance, and is up to 3 orders of magnitude faster than other methods. The particle filter is faster than real-time, and therefore is suitable in streaming applications in which the datasets can grow arbitrarily large.
Commercially-available adaptive cruise control (ACC) systems are the first step towards vehicle automation. My team devoted extensive research effors through experimentation to study the effects of the new technology in traffic streams. Specifically, we are interested in if a small perturbation of a braking event from equilibrium state would amplify or dissipate as it propagate through a platoon of vehicles.
We investigated the feasibility of a deep learning based convolutional neural network (CNN) for real-time urban mobility sensing. Specifically, we developed tools for unhelmeted cyclists detection that triggers safety alerts and methods to blur privacy sensitive information, e.g., faces and the actual plate numbers. This work proved capable of measuring traffic information (car/bike counts etc.) without building extra physical infrastructure or violating privacy of citizens. It reveals the possibility of measuring increasingly heterogeneous mobility modes that are appearing in cities, especially if we can take advantage of low cost sensor networks such as mobile camera networks.