P. Bhavsar, N. Bouaynaya, Y. Mehta, G. Rasool
CREATEs at Rowan University,
United States
Keywords: transportation asset management, IoT, machine learning, AI
Summary:
The evolving computing and communication technologies have enabled transportation practitioners and researchers to develop data-driven solutions for various transportation problems. While the transportation asset management system is one of key sub-systems (specifically for the surface transportation system), implementation of data-driven asset management system has been limited to major urban and suburban areas. One of the major deterrents is the cost of implementing and maintaining such systems. For example, every state and local governments are interested in collecting pavement condition data multiple times a year to accurately predict pavement performance. However, the current cost of operating and maintaining the existing system with a dedicated vehicle and fixed sensors is limiting several public agencies to collect data once every year or in some cases once every three years. The research team at Center for Research and Education in Advanced Transportation Engineering Systems (CREATEs) is developing an intelligent, accurate, affordable and portable solution to address this problem. Any data-driven system must convert data into meaningful information by (1) collecting/processing required data, (2) analyzing with efficient algorithms, and (3) presenting with user-specific interface. As the first step, the research team developed a prototype mobile data collection system, which is a moving Internet of Things (IoT) system that can convert any vehicle into smart data collection unit. The proof of concept of this system was presented in the previous Tech Connect conference (poster session). At present research team is working on improving mobile data collection system and developing other components of the asset management system which are data analytics and visualization. The team is working on distributed machine learning algorithms designed to specifically customize classification and identification process. Furthermore, the team will also develop prediction algorithms to predict pavement life. This is significantly important for public agencies because, replacement of one lane of one mile of the roadway can cost about $1.2 million for public agencies in the United States. Hence, accurate classification of pavement distresses and prediction of pavement life can save significant amount of time and resources. In addition, we are also integrating other sensors and object detection algorithms to develop a comprehensive and intelligent asset management system. This is the 2nd year for the asset management system and 5th year for the mobile data collection system. The New Jersey Department of Transportation (NJDOT) have sponsored the development of mobile data collection system, specifically development of pothole detection sensor. The team have filed a provisional patent in 2017 and a patent in 2018. The research team have also completed NSF I-Corp program to identify product-market fit for the product. In the current year, the research team will focus on collecting data in various weather conditions to improve existing algorithms and develop new algorithms.