Winter maintenance decision support with an Artificial Intelligence enhanced framework

Date and Time: Tuesday, May 9, 2023: 10:00 AM - 12:00 PM
Location: Keck 100

 

Lead Presenter:  Mohammad Hossein Tavakoli Dastjerdi, Researcher
Affiliation:  Ph.D. Candidate, Department of Civil, Environmental, and Geospatial Engineering, Michigan Technological University
Social Media Handle: 

Lead Presenter Biography

Mohammad Hossein Tavakoli Dastjerdi is a Ph.D. candidate in geotechnical engineering at Michigan Technological University, where he is working on his dissertation project. He has a strong background in unsaturated soil mechanics, soil improvement techniques, and image processing. Tavakoli has also taught undergraduate and graduate courses on various engineering topics at several universities.
Tavakoli's research work has resulted in several publications, including journal articles, book chapters and conference papers, on topics such as soil water retention curve, shear strength of unsaturated soils, and removal of dissolved toluene in underground water. In addition, he has also been involved in industry-focused projects, such as the evaluation of acid harm at a pharmaceutical company and the design and manufacture of new machines for soil improvement.
His interests extend beyond geotechnical engineering, as he is also proficient in machine learning, neural network models, water purification systems, chitosan biopolymer applications, special concrete admixtures, and resolving environmental issues of laboratorial apparatuses. His recent focus on artificial intelligence has allowed him to develop skills and knowledge of neural network models, and machine learning techniques. 

Co-Authors

Zhen (Leo) Liu, Corresponding Author
Associate Professor, Department of Civil, Environmental, and Geospatial Engineering,
Michigan Technological University
1400 Townsend Drive, Dillman 201F, Houghton, MI 49931
Phone: (906) 487-1826
E-mail: zhenl@mtu.edu



Dr. Liu earned his Ph.D. in Civil Engineering, with an emphasis in Geotechnical Engineering, from Case Western Reserve University in Cleveland, Ohio, in 2012. He continued working at Case as a research associate before joining the department. His teaching interests include soil mechanics, foundation engineering, Artificial Intelligence (AI), numerical simulations, and other topics in classical mechanics. His research interests are integrated for being more collaborative (multiphysics), more focused (multiscale), and more intelligent (AI). The scope covers both traditional geotechnical applications, such as unsaturated soil mechanics, geohazards, energy geotechnics and advanced geomaterials, and more interdisciplinary innovations, such as intelligent geosystems (for smart cities and cyber-physical systems), system resilience improvements, and “big data” solutions for intelligent geosystems and transportation systems. His research has many direct applications in infrastructure safety, energy resources, environment protection, and advanced materials.

Presentation Description 

Winter road maintenance aims to ensure safe and comfortable travel for road users while providing economic benefits. To achieve this, transportation data can be used in decision support systems to determine when, where, and how to apply maintenance procedures. However, current model-based decision-making tools have limitations in data utilization, decision-making improvement over time, and autonomous decision-making. This project seeks to fill these gaps by enhancing the existing model-driven winter maintenance decision-making systems with an AI-based framework called the Smart Maintenance Decision Support System (SmartMDSS.org). The system uses AI innovations like deep learning with recurrent neural networks (RNNs), deep reinforcement learning (DRL), and convolutional neural networks (CNNs) to revolutionize data utilization and decision-making. The system uses a closed-loop approach where data is collected, processed, and used to make autonomous decisions that impact the environment, whose outcomes are analyzed to validate and improve the decisions. The SmartMDSSapp uses three main AI-based processing algorithms: CNN for road surface classifications, Mask-RCNN for detecting road snow coverage percentage, and RNN for predicting road surface temperature and conditions. The system uses a Markov Decision Process (MDP), where the road engineers act as the learning agents, to make winter maintenance decisions based on reward and learn from the consequences. The app has been distributed to road engineers and is under field test in collaboration with Michigan DOT and county road agencies. The SmartMDSS app aims to fill three knowledge gaps: data utilization, decision-making improvement over time, and autonomous decision-making, and revolutionize winter road maintenance decision-making using AI.

Extended Summary

Extended Abstract

Download Extended Abstract PDF 

Presentation Video