摘要：In this paper, we study how users in a multi-channel wireless network can select frequency channels in a distributed and autonomous manner for transmitting their multimedia data such that their utilities (multimedia qualities) are maximized. In the multi-user network, the users’ transmission actions (the channel selection) are coupled by their mutual interference. However, most of the current channel selection solutions respond myopically to the aggregate interference experienced in each frequency channel. Instead, it is important to develop accurate prediction models which enable multimedia users to predict the interference from the other users and, based on the models, foresightedly optimize their decisions to maximize multimedia qualities. Such foresighted decision making is crucial for multimedia users, since they care more about long-term multimedia qualities rather than instantaneous throughput. In this paper, we discuss two multi-user interaction scenarios – non-collaborative and collaborative communication. In the non-collaborative scenario, users adapt the prediction models to maximize their own utilities. In the collaborative setting, users cooperatively maximize the same system utility (e.g. the sum of users’ utilities). A user needs to decide in which setting (collaborative or non-collaborative) it should operate, and which models it should use to predict the response of the other users, such that its own utility is maximized. We show that whether a user obtains a performance gain or experiences performance degradation depends on the prediction models adopted by the other users in different interaction settings. Therefore, to maximize its own multimedia quality, a user needs to adapt its prediction model depending on the models adopted by the other users. Hence, we propose an adaptive algorithm for the multimedia users to determine their prediction models that outperform the conventional channel selection schemes in the multi-channel wireless networks.
