2016 TransTech

UrbanITS researchers will present at UTRC conference

The following presentations will be given at the 2016 Transportation Technology Summit: Innovative Mobility Solutions organized by UTRC.

Energy-efficient Cooperative Adaptive Cruise Control of Platooning Vehicles
Weinan Gao, Zhong-Ping Jiang, Kaan Ozbay

Enormous interdisciplinary efforts have been made by automobile companies and research institutions all over the world to develop, validate and deploy autonomous vehicles aiming at assisting, ameliorating and relieving the task of driving a car. Thanks to recent advances in connected vehicle technologies and the introduction of an international standard for dedicated short range communications, the cooperative adaptive cruise control (CACC) is now realizable in the near future via vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) wireless communications. In comparison with traditional investigation on autonomous vehicles that operate in a myopic and highly localized way, CACC has the potential to further increase road safety and traffic throughput, reduce impact of transportation on the environment, and improve passengers’ comfort since it has better predictive, communication and collaborative abilities. Unfortunately, how to design safe, comfortable and optimal cooperative adaptive cruise controllers for platooning vehicles in the presence of uncertain driver behavior and complex vehicle dynamics is still a challenging issue. This research proposes a novel data-driven and model-free, optimal CACC approach based on adaptive dynamic programming approach applied to the longitudinal control of platooning vehicles with strong uncertain nonlinear vehicle dynamics and unpredictable driver behavior. The proposed approach has a strong learning capability such that the obtained approximated optimal cruise control policy can improve the safety, capacity of traffic while reducing overall fuel consumption. The proposed research is implemented in a Paramics based microscopic traffic simulation model and demonstrates its effectiveness under several mixed traffic scenarios with both human-driven vehicles and autonomous vehicles. Extensions to the fully nonlinear model will be discussed using recent developments in modern nonlinear control.

Time: 11:15AM-12:30PM, Breakout Session 3: Transportation Technology for Traffi c & Mobility Management

Location: NYIT Auditorium on Broadway, 1871 Broadway at 61st Street, New York, NY 10023


Modeling and Predicting the Frequency and Impact of Double Parking Activities in Urban Area Using Big Data
Jingqin Gao, Kaan Ozbay

Double parking is one of the key contributors to traffic congestion in dense urban areas. It routinely causes danger for cyclists, pedestrians and traffic disruptions. This study introduces a novel data-driven framework based on machine learning techniques including the LASSO, stability selection and random forest to identify influential factors and to predict the frequency of double parking events. Parking violation tickets, 311 service requests, social media information and street characteristics are utilized in the study. The random forest model achieves 85% prediction accuracy of double parking occurrences for 50 study locations in Midtown Manhattan, New York, where ground truth data is collected from recorded videos. The result also indicates that the number of hotel rooms, traffic volume, commercial usage, block length and curbside parking spaces are the top five factors contributing to double parking.

In addition, this study adopts a comprehensive modeling approach to estimate the impact of double parking with two types of models: 1) an M/M/ queueing model to estimate double parking’s effect on the average travel time; and 2) a micro-simulation model to study individual and combined effects on travel time with different levels of travel demand, double parking locations, frequency, and durations. Comparison results show that the M/M/ queueing model yields reasonably accurate predictions under uncongested traffic conditions, yet the micro-simulation model can capture the impact of additional road and traffic characteristics and provide more accurate results especially when the roadway is congested.

This study provides transportation agencies with a novel methodology to quantify the impact of double parking in a large-scale network and to predict potential double parking hotspots for better policy-making, enforcement, and management.

TIME: 9:00 AM to 4:00 PM, Tuesday, November 15, 2016

LOCATION: Poster Exhibition in Auditorium 1 Foyer, NYIT Auditorium on Broadway, 1871 Broadway at 61st Street, New York, NY 10023


Crowdsourcing incident information for disaster response using Twitter
Abdullah Kurkcu, Fan Zuo, Jingqin Gao, Ender Faruk Morgul, Kaan Ozbay

Social media data, such as Twitter data has the potential to provide valuable information for real-time traffic operations as a supplement to existing data sources such as 511 and 311. This paper compares the incident datasets of the New York metropolitan area during Hurricane Sandy from two different sources: 1) a traditional data provider that collects incident reports from multiple agencies, and 2) text information from Twitter. A text classifier, built by utilizing keywords from actual incident reports, is trained using Naïve Bayes (NB) supervised classification method to extract incident related Twitter data. The keywords are identified by Term Frequency–Inverse Document Frequency (TF-IDF) and the NB method. The filtered Twitter data is cleaned, classified into various incident types and compared geographically with that collected by the traditional data provider. The results show that Twitter could provide geolocations of specific incidents along with their intensities, durations and impact on people. Furthermore, it could also identify incidents that are not captured by traditional incident detection systems, such as gas shortage. Twitter provides an inexpensive alternative way to collect incident data in real-time and with wide geographical coverage. It can be used as an excellent supplementary data source for disaster response, incident prediction and resiliency map generation. However, based on the findings of our study, it is not recommended to use Twitter as the only data source since it is incapable of reporting some incident types, such as traffic accidents, with a high level of reliability and accuracy.

TIME: 9:00 AM to 4:00 PM, Tuesday, November 15, 2016

LOCATION: Poster Exhibition in Auditorium 1 Foyer, NYIT Auditorium on Broadway, 1871 Broadway at 61st Street, New York, NY 10023