Sensor-Driven Predictive Vehicle Maintenance and Routing Problem with Time Windows

Abstract

Advancements in sensor technology offer significant insights into vehicle conditions, unlocking new venues to enhance fleet operations. While current vehicle health management models provide accurate predictions of vehicle failures, they often fail to integrate these forecasts into operational decision-making, limiting their practical impact. This paper addresses this gap by incorporating sensor-driven failure predictions into a single-vehicle routing problem with time windows. A maintenance cost function is introduced to balance two critical trade-offs: premature maintenance, which leads to underutilization of remaining useful life, and delayed maintenance, which increases the likelihood of breakdowns. Routing problems with time windows are inherently challenging, and integrating maintenance considerations adds significantly to its computational complexity. To address this, we develop a new solution method, called the Iterative Alignment Method (IAM), building on the structural properties of the problem. IAM generates high-quality solutions even in large-size instances where Gurobi cannot find any solutions. Moreover, compared to the traditional periodic maintenance strategy, our sensor-driven approach to maintenance decisions shows improvements in operational and maintenance costs as well as in overall vehicle reliability.