AI-Assisted Optimization of Cloud Gaming Experience

AI-Assisted Optimization of Cloud Gaming Experience Cloud Gaming technology is gaining immense popularity day by day. While the idea of not paying anymore for pocket-burning hardware sounds good at first, the question immediately arises if we can at least get the same gameplay quality. The extensive source-demanding nature of Cloud Gaming technology requires more serious Quality-of-Experience (QoE) assessments on video games to preserve player satisfaction. Today, games are way more sophisticated than ever before: they are mostly composed of various game stages such as environment exploration, action-combat, dialogue with Non-Player Characters (NPC), etc. This study follows the idea that Cloud Gaming servers can allocate their resources to each player according to the game stage they are in at that moment and proposes that players' game stage can be classified by looking at instant bitrates that are being streamed to the players. The result of this study, in which Deep Reinforcement Learning (DRL) agents were used instead of the human factor that causes the QoE assessments to be cumbersome, showed us that it is possible to classify instant game stages with high accuracy by processing the corresponding instant bitrate values.