Industrial vehicles’ activities are rather heterogeneous and depend on vehicle’s type, model, and context of usage. Vehicles’ workload is usually summarized by means of standard labels, called duties. Fleet managers exploit duties to optimize productivity, minimize costs, show compliance with industrial regulations, and prevent vehicle misuses and failures. Duty levels are commonly manually enforced by manufacturers separately for each vehicle. However, since vehicles’ usage levels may fluctuate from light to heavy with variable periodicity, duty level assignment requires periodic customization. This paper addresses the problem of automatically defining the current duties of industrial vehicles and their corresponding levels by means of machine learning techniques applied to CAN bus data. In the context of industrial vehicle management, a key challenge is the lack of annotated data due to the following reasons: (1) The high variability in vehicles’ models and types makes the annotation process extremely labor-intensive. (2) Duty levels depend on the current usage context thus requiring periodic refinements. To tackle these issues, we propose a new two-stage method based on segment-wise clustering of per-vehicle CAN bus signals followed by a point-wise duty level assignment applied to the samples of each segment. Segments of per-vehicle CAN bus signals are first grouped together to discover similar vehicle usage profiles. Then, separately for each group we automatically derive the point-wise duty levels characterizing each usage profile. The proposed methodology is unsupervised, i.e., it does not require the availability of any annotated data, and is portable to arbitrary vehicles and context of usage without the need of human intervention. The effectiveness of the proposed methodology is validated on CAN bus data acquired from real vehicles of different types and models. The empirical evidence confirms its applicability to real industrial scenarios.

Learning industrial vehicles’ duty patterns: A real case / Buccafusco, Silvia; Cagliero, Luca; Megaro, Andrea; Vaccarino, Francesco; Loti, Riccardo; Salvatori, Lucia. - In: COMPUTERS IN INDUSTRY. - ISSN 0166-3615. - ELETTRONICO. - 145:103826(2023), pp. 1-15. [10.1016/j.compind.2022.103826]

Learning industrial vehicles’ duty patterns: A real case

Buccafusco, Silvia;Cagliero, Luca;Megaro, Andrea;Vaccarino, Francesco;Salvatori, Lucia
2023

Abstract

Industrial vehicles’ activities are rather heterogeneous and depend on vehicle’s type, model, and context of usage. Vehicles’ workload is usually summarized by means of standard labels, called duties. Fleet managers exploit duties to optimize productivity, minimize costs, show compliance with industrial regulations, and prevent vehicle misuses and failures. Duty levels are commonly manually enforced by manufacturers separately for each vehicle. However, since vehicles’ usage levels may fluctuate from light to heavy with variable periodicity, duty level assignment requires periodic customization. This paper addresses the problem of automatically defining the current duties of industrial vehicles and their corresponding levels by means of machine learning techniques applied to CAN bus data. In the context of industrial vehicle management, a key challenge is the lack of annotated data due to the following reasons: (1) The high variability in vehicles’ models and types makes the annotation process extremely labor-intensive. (2) Duty levels depend on the current usage context thus requiring periodic refinements. To tackle these issues, we propose a new two-stage method based on segment-wise clustering of per-vehicle CAN bus signals followed by a point-wise duty level assignment applied to the samples of each segment. Segments of per-vehicle CAN bus signals are first grouped together to discover similar vehicle usage profiles. Then, separately for each group we automatically derive the point-wise duty levels characterizing each usage profile. The proposed methodology is unsupervised, i.e., it does not require the availability of any annotated data, and is portable to arbitrary vehicles and context of usage without the need of human intervention. The effectiveness of the proposed methodology is validated on CAN bus data acquired from real vehicles of different types and models. The empirical evidence confirms its applicability to real industrial scenarios.
File in questo prodotto:
File Dimensione Formato  
Learning industrial vehicles’ duty patterns_ A real case.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 2.62 MB
Formato Adobe PDF
2.62 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2973780