Title:Analysis of a contention-based approach over 5G NR for Federated Learning in an Industrial Internet of Things scenario

Abstract:The growing interest in new applications involving co-located heterogeneous requirements, such as the Industrial Internet of Things (IIoT) paradigm, poses unprecedented challenges to the uplink wireless transmissions. Dedicated scheduling has been the fundamental approach used by mobile radio systems for uplink transmissions, where the network assigns contention-free resources to users based on buffer-related information. The usage of contention-based transmissions was discussed by the 3rd Generation Partnership Project (3GPP) as an alternative approach for reducing the uplink latency characterizing dedicated scheduling. Nevertheless, the contention-based approach was not considered for standardization in LTE due to limited performance gains. However, 5G NR introduced a different radio frame which could change the performance achievable with a contention-based framework, although this has not yet been evaluated. This paper aims to fill this gap. We present a contention-based design introduced for uplink transmissions in a 5G NR IIoT scenario. We provide an up-to-date analysis via near-product 3GPP-compliant network simulations of the achievable application-level performance with simultaneous Ultra-Reliable Low Latency Communications (URLLC) and Federated Learning (FL) traffic, where the contention-based scheme is applied to the FL traffic. The investigation also involves two separate mechanisms for handling retransmissions of lost or collided transmissions. Numerical results show that, under some conditions, the proposed contention-based design provides benefits over dedicated scheduling when considering FL upload/download times, and does not significantly degrade the performance of URLLC.