Federated System for Transport Mode Detection

Iago C. Cavalcante, Rodolfo I. Meneguette (), Renato H. Torres, Leandro Y. Mano, Vinícius P. Gonçalves, Jó Ueyama, Gustavo Pessin, Georges D. Amvame Nze and Geraldo P. Rocha Filho
Additional contact information
Iago C. Cavalcante: University of Brasília (UnB), Brasília, DF 70910-900, Brazil
Rodolfo I. Meneguette: University of São Paulo (USP), São Carlos, SP 05508-270, Brazil
Renato H. Torres: Federal University of Pará (UFPA), Belém, PA 66075-110, Brazil
Leandro Y. Mano: State University of Rio de Janeiro (UFRJ), Rio de Janeiro, RJ 20550-900, Brazil
Vinícius P. Gonçalves: University of Brasília (UnB), Brasília, DF 70910-900, Brazil
Jó Ueyama: University of São Paulo (USP), São Carlos, SP 05508-270, Brazil
Gustavo Pessin: Vale Institute of Technology (ITV), Robotics Laboratory, Ouro Preto, MG 35400-000, Brazil
Georges D. Amvame Nze: University of Brasília (UnB), Brasília, DF 70910-900, Brazil
Geraldo P. Rocha Filho: State University of Southwest Bahia (UESB), Vitória da Conquista, BA 45083-900, Brazil

Energies, 2022, vol. 15, issue 23, 1-17

Abstract: Data on transport usage is important in a wide range of areas. These data are often obtained manually through costly and inaccurate interviews. In the last decade, several researchers explored the use of smartphone sensors for the automatic detection of transport modes. However, such works have focused on developing centralized machine learning mechanisms. This centralized approach requires user data to be transferred to a central server and, therefore, does not satisfy a transport mode detection mechanism’s practical response time and privacy needs. This research presents the Federated System for Transport Mode Detection (FedTM). The main contribution of FedTM is exploring Federated Learning on transport mode detection using smartphone sensors. In FedTM, both the training and inference process is moved to the client side (smartphones), reducing response time and increasing privacy. The FedTM was designed using a Neural Network for the classification task and obtained an average accuracy of 80.6% in three transport classes (cars, buses and motorcycles). Other contributions of this work are: ( i ) The use of data collected only on the curves of the route. Such reduction in data collection is important, given that the system is decentralized and the training and inference phases take place on smartphones with less computational capacity. ( ii ) FedTM and centralized classifiers are compared with regard to execution time and detection performance. Such a comparison is important for measuring the pros and cons of using Federated Learning in the transport mode detection task.

Keywords: transport mode detection; Federated Learning; smartphone; smart cities; artificial Neural Networks (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2022
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1996-1073/15/23/9256/pdf (application/pdf)
https://www.mdpi.com/1996-1073/15/23/9256/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jeners:v:15:y:2022:i:23:p:9256-:d:995442

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

More articles in Energies from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().