TY - JOUR
T1 - Improving communication protocols in smart cities with transformers
AU - Romo-Montiel, Edgar
AU - Menchaca-Mendez, Ricardo
AU - Rivero-Angeles, Mario Eduardo
AU - Menchaca-Mendez, Rolando
N1 - Publisher Copyright:
© 2022 The Author(s)
PY - 2022/3
Y1 - 2022/3
N2 - Adaptive medium access control (MAC) protocols are essential in the context of Vehicular AdHoc Networks (Vanets) because of the rapid changes in topology caused by the high mobility of nodes. In this work, we propose an adaptive version of the Slotted-ALOHA (S-ALOHA) protocol, where the transmission probability is constantly adjusted based on estimates of the number of vehicles in the coverage area. These values are computed using deep learning models for time series prediction. One challenge for implementing this approach is that the inputs to the models are noisy since they are also estimates based on the protocol's operation itself. To address this problem, we propose a new training scheme where we add noise, similar to that produced by the protocol's operation, to the inputs of the training examples as a form of regularization. Our experiments show that the regularized models perform close to the theoretical optimal where the number of vehicles in the area is always known.
AB - Adaptive medium access control (MAC) protocols are essential in the context of Vehicular AdHoc Networks (Vanets) because of the rapid changes in topology caused by the high mobility of nodes. In this work, we propose an adaptive version of the Slotted-ALOHA (S-ALOHA) protocol, where the transmission probability is constantly adjusted based on estimates of the number of vehicles in the coverage area. These values are computed using deep learning models for time series prediction. One challenge for implementing this approach is that the inputs to the models are noisy since they are also estimates based on the protocol's operation itself. To address this problem, we propose a new training scheme where we add noise, similar to that produced by the protocol's operation, to the inputs of the training examples as a form of regularization. Our experiments show that the regularized models perform close to the theoretical optimal where the number of vehicles in the area is always known.
KW - ALOHA
KW - LSTM
KW - Machine learning
KW - Transformers
KW - Vanets
UR - http://www.scopus.com/inward/record.url?scp=85125517173&partnerID=8YFLogxK
U2 - 10.1016/j.icte.2022.02.006
DO - 10.1016/j.icte.2022.02.006
M3 - Artículo de revisión
SN - 2405-9595
VL - 8
SP - 50
EP - 55
JO - ICT Express
JF - ICT Express
IS - 1
ER -