On Designing Protocols for Noncooperative, Multiradio Channel Assignment in Multiple Collision Domains

Abstract

Abstract-Channel assignment is a crucial problem for wireless networks, especially for non-cooperative wireless networks, in which nodes are selfish. While there have been a few studies of non-cooperative, multi-radio channel assignment, most existing studies are restricted to single collision domains only. In this paper, we study the design of incentive-compatible protocols for non-cooperative, multiradio channel assignment in multiple collision domains.First, we show the necessity of designing incentivecompatible protocols for this problem. Specifically, we show that, if no incentive-compatible protocol is deployed, Nash Equilibria (NEs) may have undesired properties, such as Pareto suboptimality and low throughput.In order to prevent the system from converging to the NEs with undesired properties, we propose an incentivecompatible protocol for channel assignment in multiple collision domains. We rigorously show that our protocol guarantees that the system converges to NEs that are Pareto-optimal and have the maximum system-wide throughput. Our simulation results also verify that our protocols are effective in ensuring that the system converges to the desired NEs.

Multi-Radio Multi-Channel Allocation in Competitive Wireless Ad Hoc Networks

“…Recently, game theory has been considered to solve the multi-channel allocation problem in competitive wireless networks [7][8][9][10][11][12][13][14]. Felegyhazi et al first introduced the strategic game model to study the non-cooperative behavior of channel allocation in competitive networks [7].…”

“…While, they also considered all senders and relaying nodes were in a single collision domain, which is unsuitable for the long path. Chen et al went one step further to study the problem of non-cooperative multi-radio channel assignment in multiple collision domains [12] [13]. They designed incentive-compatible protocols to guarantee complete fair channel assignment and to avoid Pareto suboptimality in multiple collision domains.…”