Delay-Sensitive Cross-Layer Optimization for Wireless Systems (click here)

5G wireless networks brings a lot of new applications as well as challenges. In a nut shell, there is 1000X increase in capacity demand, 1000X increase in connectivity and 100X reduction of latency. These are very challenging demands which cannot be addressed by one single solution. A portfolio of solutions can be divided into (1) New Physical Layer Technologies, (2) New Radio Resource Management and (3) New Architecture.

From very high level, if one looks at the Shannon’s capacity formula, there are only a limited number of parameters that can contribute to capacity increase. (a) Increasing number of spatial channels. There are various research topics on new techniques that can contribute directly and indirectly to increasing the number of spatial channels. For example, in massive MIMO, more spatial channels are created with massively many antennas at the base station using MIMO techniques. In small cell networks or heterogeneous networks, there are higher spatial reuse in a cellular network, which also contribute indirectly to more spatial channels.

Another possibility is to (b) increase the bandwidth in the Shannon’s capacity formula. There are various research topics that contribute to increase in bandwidth. For example, carrier aggregation techniques, cognitive radio techniques, unlicensed LTE systems and mmWave technologies can contribute to higher bandwidth.

A third possibility is to (c) increase the SINR in the Shannon’s capacity formula. Research topics such as advanced error correction codes, interference mitigation, Interference alignment, coordinated beamforming in cellular network, massive MIMO , CoMP (networked MIMO) and advanced radio resource management (RRM) all contribute to improving the SINR.

(I.1) New Physical Layer Technologies for 5G Wireless Systems

Typical radio resource management attempts to maximize the PHY performance regardless of the applications. The wireless communication network is to provide a transparent pipe without worrying about what specific application is using the pipe. This sometimes results in suboptimal designs for some niche applications. One can view the design of radio resource management as an optimization problem. For instance, the notion of “cross-layer” may refer to “crossing layers” in optimization variables such as the PHY variables: (MIMO precoders, transmit powers, rate adaptation), the MAC variables (user scheduling, subband allocation) and the Network Layer Variables (multi-hop route, admission control). The notion of cross-layer may also refer to “crossing-layers” in optimization objective(s) and constraint(s). For example, one can optimize the PHY performance (such as the sum capacity, throughput, proportional fairness) or the MAC performance (such as queue stability, average delay, delay jitter at the MAC queues) or most preferrable, the end-to-end application performance (such as video streaming playback interruption probability or the networked control stability.). To simplify the problems, one usually focus on optimizing the PHY performance, believing that the resulting end-to-end application will automatically perform good. Unfortunately, this turns out to be not the case for some applications. For example, in video streaming applications, it is not the bit rate or average delay that matters. It is the video playback interruption probability or buffer overflow probability that matters. Another example is the machine-to-machine communications for networked control applications. In this application, it is again not the bit rate nor average delay that matters. The goal is for sensors (machine nodes) to measure the system state and provide state feedback to the controller (closed-loop control) so as to stabilize an unstable dynamic system (such as a chemical plant). In this application, the RRM should be aware of the underlying application so as to optimize the wireless resource to complete the task (stabilization of industrial plant).

(I.2) New Radio Resource Management for 5G Wireless Systems

In addition to advanced PHY and RRM technologies, it is also important to have a holistic design to support the challenging demands for future 5G wireless systems. Several new architectures are explored to support future 5G wireless networks. One example is Cloud-Radio Access Networks (C-RAN). Different from traditional cellular networks, each base station in C-RAN is called a “remote radio head” (RRH) where it only performs simple RF processing. The IF or baseband samples are then quantized and transmit to the “baseband cloud” (BB-Cloud) for centralized base band processing. As such, it offers the advantage of small cell network and support high mobility because the notion of “cell” is not defined by the RRH coverage (but rather defined by the CoMP clustering boundary in the BB cloud). It also offers a lot of other advantages such as reducing interference (bring the transmitter closer to the mobile), more efficient interference mitigation (due to centralized processing) as well as virtualization of base station (soft base station). While the C-RAN offers a lot of potential new opportunities, there are also a lot of practical challenges that need to be addressed. Another new architecture we explore is called the PHY caching in wireless networks. Traditional wireless networks are designed to deliver raw information bits from the source to the destination. However, in practical scenarios, the wireless network should be designed to deliver contents. There is a subtle difference between random bits and content --- contents are cachable. A packet transmitted by this base station may be requested by other users in the near future. So, if the base station cache this packet, then the demand can be addressed from the cache without consuming the backhaul bandwidth. While cache has been widely explored in the internet and content distribution network (CDN) or peer-to-peer networks (P2P), the proposed PHY caching (our group is the first to propose PHY caching to induce MIMO cooperation opportunities)  is fundamentally different from these conventional cache. In conventional cache, the benefit comes from reducing the number of hops from the source to the consumer as well as load-balancing. The underlying PHY of the links as well as the network topology is the same regardless of the cache state (hit or not hit). However, in the proposed PHY caching, the role of cache is to induce a favorable PHY topology (e.g. from unfavorable interference channel to more favorable broadcast channels). In other words, the underlaying PHY of the wireless network changes dynamically depending on the random cache state at the BS. As such, the PHY cacheing provide a new perspective of designing wireless networks. A lot of wireless network technologies (which was regarded as providing only marginal performance benefits) such as relay and adhoc network will offer substantial gains by introducing PHY caching.

(I.3) New Architecture for 5G Wireless Systems