On a game theoretic approach to capacity maximization in wireless networks

Sleeping Experts in Wireless Networks

“…This approach extends related works on capacity maximization on a single channel with uniform powers in SINR [1,9] and Rayleigh-fading models [5], for which no-external regret algorithms are known to converge to constant-factor approximations [2,5,6]. Closest to our approach is our companion paper [4], in which we introduce a general framework based on the LP technique to study no-external regret learning with adversarial jamming on a single channel.…”

“…For static availability, it is easy to see that previous results on no-external regret learning in [2,5] directly extend to multiple channels. When we consider varying availabilities, however, multiple channels represent a significant complication.…”

“…If all links use the same uniform power for transmission, this results in Cindependence with a constant C. This was proven by Ásgeirsson and Mitra [2]. Using affectance it is straightforward to construct the corresponding conflict graph by simply setting weights b u (v) = a (u, v).…”

“…We define the following notion of C-independence inspired by [2] as a key parameter to identify the connection between the specific interference model and the performance of our algorithm.…”

“…Throughout this paper we define the utility of a link i as follows. This utility function was already used for a single channel case where the channel is always available by Andrews and Dinitz [1] and later by Ásgeirsson and Mitra [2].…”