Fuel-based trip-level behavior optimization employing metaheuristics

Driving behavior, primarily speed and acceleration, has a significant impact on fuel consumption. As a result, fuel economy can be severely altered by making behavior-level interventions in the driving habits. This paper presents a framework for optimizing driving behavior to minimize estimated fuel consumption while preserving selected features of the recorded trip through explicitly defined trip-level constraints. To achieve this, an analytical study was performed with the use of a metaheuristic algorithm and a sampling algorithm in a novel framework. The algorithms used were Ant Colony Optimization as the primary optimization algorithm, and Latin Hypercube Sampling as a space-filling candidate-generation algorithm. The objective function integrates longitudinal motion parameters such as speed, harsh deceleration and acceleration, etc. as significant fuel consumption indicators through a physics-based algorithm. The proposed framework is applied to a large-scale naturalistic dataset that incorporates almost 33,000 trips recorded in Athens, Greece, between March and May 2024. The framework was evaluated on a sample of 100 trips randomly selected from the dataset recorded. Under the modeled assumptions, the results suggest that optimizing individual trip speed profiles may improve fuel economy and reduce the environmental impact of urban travel. These profiles represent theoretical outcomes and are not ready for use as driving recommendations.