MUTE: Sounding inhibition for MU-MIMO WLANs

Abstract

In this paper, we present the design, implementa- tion, and evaluation of the novel downlink Multi-User MIMO sounding protocol called MUTE. Our protocol decouples the sounding set selection used to collect Channel State Information (CSI), from the transmission set selection in order to minimize or even eliminate the overhead associated with sounding, while maximizing user selection performance. To this end, MUTE exploits channel statistics to all the different users to predict whether a particular user's channel will remain sufficiently stable, thereby allowing the access point to preclude channel sounding before a MU-MIMO transmission. We show that in indoor WLANs, MUTE can reduce sounding overhead by close to 73% under certain conditions while minimizing rate performance losses due to inaccurate channel estimation. Zero-Forcing Multi-User-MIMO beamforming systems (ZFBF MU-MIMO) rely on channel sounding to provide the beamformer or Access Point (AP) with Channel State Information (CSI) about each beamformee or user. This is necessary to generate the steering beam weights required to perform the zero-forcing precoding prior to a beamformed transmission (13). Additionally, it is advantageous to acquire CSI from all associated users in order to maximize user diversity, 1 thereby improving the user selection process by increasing the likelihood of finding beamformees with orthog- onal or semi-orthogonal channel vectors. This can lead to complete suppression of interference between the different data streams serving the different beamformees and therefore to rate maximization at every transmission. To this end, the beamformer can acquire channel estimates from all potential beamformees before every packet transmis- sion. This provides the AP with accurate, up-to-date CSI about all users to be served, hence improving the performance of the precoding scheme. That is, having the most updated CSI for all users allows the AP to find the optimal user grouping strategy at every transmission. Unfortunately, the overhead required for CSI acquisition is directly proportional to the number of users to be sounded as well as the frequency with which this process takes place. Therefore, in a practical system, the beamformer should find a balance between sounding frequency and CSI accuracy, in the interest of minimizing sounding overhead. In this paper we propose a multi-user zero-forcing beam- forming sounding protocol that addresses the issue of overhead associated with channel sounding, with the goal of eliminating 1 We define user diversity as the accommodation of a finite set of users with distinct channel characteristics. it temporarily based on channel stability. We name our protocol MUTE which stands for Multi-User Transmission Enhancer. In the best case, in the presence of users with stable channels, MUTE will invoke a MU-MIMO transmission without any im- mediately preceding channel sounding, thereby vastly reducing overhead and correspondingly increasing transmission air time and throughput. Nonetheless, MU-MIMO is very sensitive to the accuracy of CSI, specially as the number of concurrent streams increases. Therefore, MUTE strives to find a balance between CSI degradation and sounding suppression. We argue that the decoupling of the sounding selection pro- cedure from the transmission user selection procedure provides the flexibility to choose whether to sound a particular user or not, independently from the set of them to be served in the next ZFBF transmission. This in turn decreases overhead associated with sounding by exploiting the presence of users with stable channels, while independently providing sufficiently accurate information to the AP about channel statistics of associated users. Then, based on this information, the AP can select the combination of users that maximizes an objective function such as achievable rate or a fairness criteria, for example. This is in contrast to existing MU-MIMO implementations where the set of sounded users is the same as the set of users to be served next (4), (7), (12). Furthermore, two of the major strengths of MUTE are interoperability with IEEE 802.11ac (3) devices as well as the fact that it can operate independently of the scheduler implemented.