Comparative Analysis of Mass Estimation Methods for Heavy-Duty Vehicles under Varying Payload Conditions

<div class="section abstract"><div class="htmlview paragraph">Accurate mass estimation is essential for commercial heavy-duty vehicles (HDVs) because fluctuating payloads significantly impact energy consumption. Precise vehicle mass estimates enhance the accuracy of energy consumption models, leading to more effective energy management systems and performance optimization strategies. For example, improved energy estimates can lead to more optimized routing and refueling schedules, improving operational efficiency and reducing costs. For electric HDVs, accurate mass estimates are crucial for battery sizing, range prediction, and optimized charge scheduling. While direct mass measurements may be obtained through external weight-in-motion or specialized onboard weighing systems, this paper focuses on methods that use data from Controller Area Network systems for alternative real-time predictions. The challenge lies in identifying a method that performs well under the highly variable and often sparse data conditions typical of HDV driving datasets. Three mass estimation approaches are evaluated in this work: the k-nearest neighbors (kNN) regressor, the recursive least squares (RLS) method, and a feed-forward neural network (FFNN). Data-driven methods like kNN regressors perform well with limited data but are computationally intensive during inference. Vehicle model-based methods, like the RLS method, offer high explainability but rely on accurate parameterization and high-resolution data for critical calculations. FFNNs, though powerful, require large datasets for effective generalization and lack interpretability. The comparative analysis assesses each method for accuracy, robustness to sparse data, and suitability for real-time application. Telematics data from Class 8 trucks with varying payloads are used for validation, with performance measured using mean absolute percentage error (MAPE). The kNN regressor achieved the highest accuracy, followed by the RLS method and the FFNN. The RLS method was the fastest, while the FFNN required the least memory for storage and inference. Both kNN and FFNN maintained robust performance under sparse data conditions, unlike RLS which required continuous data streams for accurate estimation.</div></div>