Objective QoE Models for Cloud-Based First Person Shooter Game over Mobile Networks

Mobile cloud gaming (MCG) lets users play cloud games (CG) on mobile devices anywhere via mobile networks. However, the stochastic nature of network quality of service (QoS) can result in varying user quality of experience (QoE). Understanding, modeling, and predicting the impact of mobile networks' QoS on users' QoE is crucial. This helps stakeholders optimize networks, and game developers efficiently create cloud-hosted games provisioned over mobile networks. This paper investigates the impact of QoS on users' QoE and proposes, develops and validates novel models for predicting QoE for MCG in mobile networks using realistic subjective tests. In particular, we propose and develop three QoE models using multiple, polynomial, and non-linear regression. Our results validate that multiple regression (with R
²
=0.79, RMSE=0.45) can model complex relationships between QoS factors that impact QoE. Multiple polynomial regression achieved the overall fit with (R
²
=0.94, RMSE=0.24). Lastly, the non-linear model achieved a good RMSE of 0.24. To select the best model out of the three, we applied the F-test and determined that polynomial regression had the best statistical fit.