This workshop focuses on artificial intelligence and machine learning techniques for turning high volumes of fast-moving and diverse transportation data into useful information. Advanced data survey and analytics techniques and their applications to emerging big data sets (e.g., trajectory, social media, mobile sensing, and connected and automated vehicles) and the potential of big data in supporting needs such as performance measurement, forecasting, and real-time control are explored.
This talk is about the interactions between two well-known manifestations of human behavior: cities and language. The main angle will be the transportation system, and I’ll explore how human communication data can improve predictability (and understandability) of both demand and supply aspects. Such opportunity has been generally neglected, differently to other recent trends based on other datasets (e.g. telecom data, crowdsourcing, smartcards). I will argue that human communication brings the semantic enrichment needed to complement these other dataset and to better comprehend our cities. The talk will include examples from recent work in Singapore, involving traffic prediction, public transport planning and activity-based modeling.
Advances in IoT and Connected Car technologies have highlighted the importance of maps and networks to the future of mobility services. Currently devices are collecting and delivering location data at a scale greater than ever. Concurrently, cloud computing is offering scaled computation with tools that reduce the resources necessary to implement parallel computing solutions that can address the scale of the data. While the scale has been changed by technology, the underlying scientific methods must still remain sound. Transforming the data into knowledge will become increasingly challenging and will require frameworks, such as mapping frameworks, that bound inferences and can provide confidence in the transformations. Beyond the challenges with data analytics, future applications will need to consider the forming ecosystem around these devices and services as well as the computational capabilities of the devices and networks, in addition to the cloud.
In this talk, I will present a general overview of the Big Data paradigm with special emphasis on the data science components. I will also talk about the future applications which will include third and fourth generation analysis—Deep Learning and Adaptive Learning. I will cover both the challenges and perspectives of the third generation analysis. Some of the examples to be covered include: a) the relationship between rail defects and track geometry, b) rail surface damage monitoring, and c) multiple wheel load impact detection systems data.
The talk will include the following three topics in social media data analysis. They are related to data mining and machine learning.
(1). Predict transit passenger flow with social media data.
(2). Study and explain the traffic surge with social media data.
(3). Identify on-site traffic accidents using both traffic and social media data.
We used a variety of ML techniques: k-means clustering, classification with a Convex Optimization with Hybrid loss function, filtering, peak-detection, topic modeling, logistic regression, etc.
Transportation data generated by social networks, sensors, and mobile devices has made us move into the era of “big data”. A great challenge in this era is how to extract useful information from high volumes of real-time and historical transportation data for performance evaluation of transportation systems, congestion reduction, prediction of future transportation needs, and real-time traffic management of existing transportation infrastructure.
Los Angeles County Metropolitan Transportation Authority (LACMTA or LA Metro), in working with other regional transportation agencies (including Caltrans, California Highway Patrol, and local cities), has established the Regional Integration of Intelligent Transportation Systems (RIITS) program and a Big Data system that collects all the real-time and archived transportation data within Southern California. The data contains information about buses from LA Metro and other regional bus providers, traffic signal timing, freeway operation, CHP incidents, ramp metering, changeable message sign, CCTV, etc.
This presentation aims to provide an overview of this Big Data system and describes a set of case studies that demonstrate how to turn RIITS data into insightful information for transportation system analysis. The case studies include traffic flow theory verification, freeway weaving analysis, ramp meeting performance evaluation (or overspill study), and I-405 closure analysis. This presentation also outlines future tasks that will use the Big Data system for assessment of regional transportation systems, mobility assessment of the region, and others.
Connected vehicle technology is expected to produce massive amount of data. The size and frequency of data generated from Connected Vehicles poses interesting challenges to Data Scientists. To yield the benefits of connected vehicle technology the data needs to consumed and used in real-time by applications to yield safety, mobility and environmental benefits. This presentation will discuss challenges associated with evaluating the benefits of connected vehicle applications and the advancements in modeling tools and techniques to support the evaluation.
This presentation will provide background on the SHRP2 Naturalistic Driving Study and the tools that are being used to work with this rich data set. This includes the establishment of the Safety Training and Analysis Center, which provides a secure data enclave for FHWA and state-based researchers to work with the data. I’ll also describe recent Big Data projects that explore the use of existing data sets and SHRP2 data to determine crash risks and effective countermeasures.
FHWA has demonstrated open research data access through the availability of 13 research data environments available on the new Research Data Exchange (RDE). This presentation will summarize current RDE capabilities, data environments, and lessons learned in initiating this research data open access portal.