Work

QoE Assessment for Cloud-streamed Games based on Subjective Tests

Year: 2026
Type: dissertation
Abstract: Interest in Cloud Gaming (CG) has been increasing over the years. CG is when games are rendered and streamed from a distant server to a local computer. This offers offloading for end users' hardware, and thus, they do not require high-end hardware to play demanding games on the highest settings and higher frame rates. However, the end users require a monitor that is sufficiently capable of handling the streamed data, and more importantly, a stable network. Research so far has indicated that the Quality of Experience (QoE) of the end users is heavily reliant on the Quality of Service (QoS) given by the network providers. QoE is a subjective measurement of how good or bad users perceive an experience. QoS is an objective measurement of how well a service is provided; in this paper, the QoS metrics are variables of a network. The QoS metrics covered are Round-Trip Time (RTT), Bitrate, Random Jitter (RJ), Packet Loss (PL), and Burst Jitter (BJ). The goal of this research is to further understand how different QoS metrics impact end users' QoE in CG. To do so, a subjective experiment with two different games and 15 network conditions was held. The two games tested was Counter-Strike 2 (CS2), a First Person Shooter (FPS), and Terraria, a open world two-dimensional side scroller. Except for the baseline condition, there were 2 RTT-, 4 Bitrate-, 3 RJ-, 3 combined RTT/PL-, and 2 BJ conditions. Each test was conducted for 90 seconds, and the participant rated their perceived overall experience, video quality, control, responsiveness, and willingness to continue playing under the condition. The streaming software pair Moonlight and Sunshine was utilized for client and host, respectively. The software ALTRUIST was used for automation, application of conditions, and linking all devices together. To invite participants, papers with experiment details were handed out on campus, and a verbal summarization of the experiment was given. Papers were also put on campus billboards. 31 individuals participated in the experiment, of which 7 were female, and 23 were male. The best three conditions for CS2, excluding the baseline was Bitrate(56 Mbps), RJ(27 ms, 1 std), and Bitrate(10 Mbps) with the QoE Mean Opinion Score (MOS) of 5.6 , 5.48, and 5.16, respectively. For Terraria, the three best were RTT(27 ms), Bitrate(56 Mbps), and Bitrate(10 Mbps), with the QoE MOS of 5.52, 5.51, and 5.46, in the order given. The worst three conditions for CS2 was RTT(402 ms), RJ(27 ms,3 std and PL(102 ms, 2%), with the QoE MOS of 1.97, 2.44, and 2,92, respectively. For Terraria, the worst three were RJ(27 ms, 3 std), RTT(402 ms), and BJ(700 ms BJ), with the QoE MOS of 3.09, 3.21, and 3.96, in the order given. Terraria was more resistant compared to CS2 in most conditions, with the difference being more noticeable in RTT (402 ms), Bitrate (0.5 Mbps), RJ (27 ms, 3 std), and all PL conditions. The results of the experiment indicates that the QoS requirements for an acceptable QoE within CG utilizing Moonlight and Sunshine is: a RTT below 102 ms; a Bitrate above 5 Mbps; a RJ(27 ms) with a std of 2 or lower; a 2% PL is not acceptable to the majority in CS2 and playable in Terraria when RTT is 52 ms or below; a BJ of 200 ms or lower in 15 second intervals;
Author: Myhr, Tore
Language: English
Open Access status: green
OpenAlex ID: W7135163136
Cites: 11
Cited by:
Related to: 10