A parametric quality model to evaluate the performance of tele-operated driving services over 5G networks

Abstract

We have developed a parametric model to quantify the Key Quality Indicators which affect video-based Tele-operated Driving (ToD) over a mobile network, as well as their relationship with the network Key Performance Indicators. This model can be easily used to specify Quality of Service policies (e.g. through network slicing) that guarantee the required conditions for remote driving on specific areas. We have used our model to validate the feasibility of deploying remote-assisted driving in different real networks, both from current 4G deployments and from pre-commercial and commercial 5G pilots. Our results show that some ToD services (supervision and, up to some point, parking) may be feasible with high-end existing 5G networks. However, full remote driving requires some improvements in the system, particularly to reduce end-to-end latency, increase uplink performance, and minimize service losses. Both the model and its results will be used in the framework of European Union H2020 project 5G-MOBIX to deploy a ToD proof-of-concept in the cross-border corridor between Spain and Portugal.

A Survey on Remote Operation of Road Vehicles

“…Finally, authors in [68] evaluate the performance of 4G and 5G networks in remote driving contexts. Using measurements obtained in a cross-border corridor between Spain and Portugal, authors observe that 5G networks are not optimized for up-link contributions (e.g., to send sensor information or video streams), which might impair remote driving.…”

“…However, considering the potential for driving automation systems to operate correctly within an ODD, it might not be required to have RORV available at all times and in all roads, and if so, not under the stringent network requirements that remote driving and fallback driving have. As it was the conclusion of works measuring latency for remote driving applications, even 5G networks would require optimization to allow remote driving [68], but remote assistance -supported by high level ADS -could be bolstered by currently deployed network technologies ( [76], [77], [78], [79]). Also, the concept of whitelisting [56] is an approach that addresses the fact that high-performing network technologies are mostly available in urban areas, where remote driving might be more on demand than on highways.…”

“…On the one hand, Phantom Auto [12] explicitly highlights its potential to enable fleet managers to recruit drivers that work remotely, and thus mitigate regional or even national labor shortages, and by extension access qualified labor at a lower cost. On the other hand, [68] studies remote driving in an international corridor between Spain and Portugal, where not only there is a change in countries and operators (although roaming is possible, it does involve certain performance compromises, as analyzed in [112]), but the two countries also use two different languages. These factors affect the viability of remote operation as a business, since there is an intrinsic cost of training a roster of remote operators to be able to deal with regulations from different countries and regions (e.g., differences in language, driving sides, and signaling standards) potentially within the same workday or even simultaneously.…”