Customized Machine Learning for Safer and Smarter Transportation Applications
Yinhai Wang
Civil and Environmental Engineering, University of Washington
ABSTRACT: Recent advances in sensing, networking, and computing technologies, have led more and more cities to launch smart city plans to improve quality of life, sustainability, efficiency, and productivity. Sensor networks are essential for smart cities, and many new transportation-related data and computational resources are expected within the Smart Cities environment. However, classic traffic analysis methods are not designed to analyze and process the big-data sets generated from sensors. To take full advantage of these data sets, new methods and tools are needed. Machine learning methods have been increasingly utilized in transportation applications, but most have been developed in other fields and might not fit well. Customizations of conventional machine learning methods are highly desirable. Wang introduces a couple research efforts made at the University of Washington Smart Transportation Applications and Research Laboratory (STAR Lab) that produced customized machine learning methods for safer and smarter transportation applications. The superb performance of these customized machine learning methods clearly indicates the value of such artificial intelligence methods.