The ubiquity and flexibility of wireless access to the Internet played a very significant role in the tremendous growth in mobile devices such as smartphones, tablet PCs, and laptops over the past years. As a consequence, a larger and larger fraction of Internet traffic is delivered wirelessly. How to deal with this growth is one of the most important challenges for future wireless networks. State-of-the-art wireless communication already operates close to Shannon capacity and the only viable option to further increase data rates is to increase the communication bandwidth. Very high bandwidth channels are only available in the extremely high frequency part of the radio spectrum, the millimeter wave band (mm-wave). Upcoming communication technologies, such as the IEEE 802.11ad standard operating at 60GHz, are already starting to exploit this part of the radio spectrum. However, this part of the spectrum suffers from high attenuation and signal absorption, restricting communication primarily to line-of-sight (LOS) scenarios.

This in turn requires a radical rethinking of wireless networking in the mm-wave band from 30 to 300GHz. In analogy to the evolution of wired Ethernet from a shared medium to a fully switched network, we envision that future wireless networks will consist of many highly directional LOS channels for communication between access points (APs) and end devices. Such an environment is extremely dynamic and channels may appear and disappear over very short time intervals, in particular for mobile devices when persons move about in their vicinity. At the same time, such channels experience very little interference and resources (time, frequency, signal processing, etc.) that would otherwise be used to handle interference can now be used to further increase achievable data rates between sender and receiver. To provide sufficiently many LOS channels, APs may have to be deployed ubiquitously and may vastly outnumber mobile devices.

We propose to design and build a wireless network architecture that maintains a number of directional LOS channels between several APs and (mobile) end devices through transmit beamforming and beam steering. Data is transmitted simultaneously via all of these channels. An end device uses multiple antennas to receive and decode several such data streams, and the higher the number of received streams, the higher the data rate achieved at the receiver. The main complexity of the design lies in the selection of AP s as well as the beamforming directions of their antennas, given the large number of end devices that future wireless networks will have to support. To aid and speed up this decision process, the system maintains an up-to-date map of the radio environment and learns likely sequences of beamforming patterns and succession of APs. This further allows to intelligently switch off APs to improve energy efficiency. We believe that such a design is the key element for the scalability of future wireless networks.