Session: Track 1-5B: Track Safety

Paper Number: 129728

129728 - Robust Artificial Intelligence for Reasoning and Decision-Making in Mitigating Ground Hazards Along Railroads 

Every year in the United States, geohazards threaten lives, infrastructure, and livelihoods, resulting in billions of dollars in damage. However, they vary in magnitude and complexity, where one geohazard can be responsible for the generation of others and, or serve as a potential trigger for other geohazards. The United States has the most extensive rail transport network in the world, with tracks traversing some of the most challenging and complex environments and human development that could cause rapid onset of ground hazards such as landslides, subsidence, and flooding. The railroad network is the lifeblood of United States freight transportation, accounting for more than 25% of all freight. Ground hazards are inevitable; unfortunately, mostly reactionary measures are currently in place when they happen. To solve this problem, mitigation measures can be adopted using remotely sensed data from drones, satellites, ground penetrating radar, and traditional condition monitoring data. Using these datasets, we can detect, predict, and characterize the ground hazard susceptibility of the tracks. Using these analyses and models, we can build a robust AI system integrated into a decision support system (DSS) to help decide which parts of the track are of high, medium, or low susceptibility to ground hazards.

Artificial intelligence (AI) is a system capable of performing tasks associated with intelligent beings. A robust AI can apply its knowledge systematically to reason flexibly and dynamically to approximate real-world conditions.  Using a wide array of information and data gathered and recent advances in AI research provides an opportunity to implement a robust AI as part of a DSS that is being developed to help identify and reduce the risk of ground hazards along railroad tracks.

Large remotely sensed data and text files generated from traditional condition monitoring have been gathered on ground hazards and used to characterize the railroads in our areas of interest. In addition, change and anomaly detection, and predictive flood models have been developed using a combination of different remotely sensed data. These models' inferences, facts, and information about the railroad provide a wealth of knowledge that can be used as a deeply structured knowledge base for creating a robust AI system.

This paper discusses the constraints in implementing a robust AI for ground hazard mitigation, the data, information, and model inferences relevant to creating the knowledge base for geohazards along railroad tracks, and the design of the components of the AI system being implemented. In conclusion, a robust AI system as part of a DSS is being developed to proactively locate, monitor, and mitigate surface and subsurface hazards along the right of way.

Presenting Author: Abdul-Rashid Zakaria The University of Mississippi

Presenting Author Biography: Abdul-Rashid Zakaria, EIT is a Ph.D. Computer Science student at the University of Mississippi focusing on artificial intelligence and remote sensing. He is a research assistant on the Federal Railway Administration project: An Integrated and Automated Decision Support System for Ground Hazard Risk Mitigation for Railways using Remote Sensing and Traditional Condition Monitoring Data. Zakaria holds a BSc. in Geological Engineering from Kwame Nkrumah University of Science and Technology, Ghana, and an MS in Civil Engineering from Villanova University, Pennsylvania.

Authors:

Abdul-Rashid Zakaria The University of Mississippi
Thomas Oommen The University of Mississippi
Pasi Lautala Michigan Technological University
Colin Brooks Michigan Tech Research Institute