Adapting Human and Artificial Intelligence for Safe, Sustainable, and Adaptive Operations of Commercial Vehicle Fleets

The proposed project intends to enable “Adaptive Human and Artificial Intelligence (AI) for Lean Operational Safety of Commercial Vehicles Fleets,” which allows humans and AI to collaborate in proactively reducing idling time, waste, and carbon emissions while keeping or improving mobility and safety of various fleets with diverse deterioration trends in various usage environments. Scientific and technical challenges to this vision and the practical use of vehicle fleet digital twins for fleet managers and stakeholders (e.g., policymakers) include 1) how to find the best ways of integrating human experiences with data-driven predictions of the digital twins in the safe and efficient management of diverse vehicle fleets; 2) how to make the truck operational risk quantification and inspection-maintenance suggestions of the digital twins explainable for effective human-machine collaboration in reducing waste while keeping fleet-level safety and demand satisfaction.

Collaborating with two industry partners (Safety Emissions Solutions and Truck-Lite) in the past two years, we used inspection reports and telematics data to create and update data-driven simulations of vehicle deterioration processes, forming "deterioration digital twins" of commercial vehicle fleets that can support "what-if" safety-mobility-waste tradeoff analyses of fleet operators. Given a fleet and real-time data, self-updating vehicle fleet digital twins can help fleet managers decide on high-risk vehicles for proactive maintenance planning with high precision while reducing the number of inspections and downtime. The next goal is to work with three industry partners (two deployment partners, Safety Emissions Solutions and Truck-Lite, and one equity partner, Trade Institute of Pittsburgh) to enhance the practical use of these digital twins to influence policymakers regarding technology adaption and safety programs at local and national levels, and support fleet management decision-making and equipment workforce training.

The technical approach is to:

  1. analyze the reliability of the predictions of the vehicle deterioration digital twins on various fleets operated by the industry partners
  2. integrate the causal analysis of crashes vehicle deterioration (focusing on how autonomous vehicles have different brake and tire deterioration rates than human-driving vehicles), and human behaviors in interacting with computers for identifying high-risk decision options and safety-mobility-waste tradeoffs
  3. work with an equity and workforce development partner (Trade Institute of Pittsburgh) to establish training programs for commercial vehicle inspectors, drivers, and equipment managers

Specifically, the implementation partners have access to:

  1. millions of periodically scheduled vehicle inspection records (brakes, tires, lights, and many other safety-related vehicle components under regulation) of heavy-duty tractors and trailers in Western Pennsylvania (accumulated in the past 20 years by Safety Emissions Solutions)
  2. telematics data that provide real-time monitoring of safety-critical vehicle components for several vehicles owned by Truck-Lite
  3. inspection and crash records obtained from Federal Motor Carrier Safety Administration (FMCSA), and emission data to be obtained from PennDOT to assess “lean operations” strategies that maintain the necessary redundancies of vehicles without wasting resources and generating unnecessary carbon emissions

We plan to use these data sources to develop methods that automatically generate “lean safety” operational suggestions adaptive to real-time contexts that keep the minimum redundancies for commercial vehicle fleets' operational safety and sustainability.