Skip to main content

Towards an efficient mobile Internet

Final Report Summary - MOBILENET (Towards an efficient mobile Internet)

It is expected that, very soon, the Internet will not only connect billions of mobile device users, but also objects such as household appliances or components of transportation vehicles, building a genuine Internet of Things. This places high demands on the communications infrastructure and on the mobile devices. Specifically, future networks should support heterogeneous mobile, wired, and wireless network technologies; be dynamic, self-organizing, and robust; and at the same time offer high data rates and use resources, such as bandwidth and energy, in the most efficient way. Future mobile devices should be small and have low energy-consumption while offering high functionality.

To explore how to use the resources in future communication networks in the most efficient way, the aim of this project was to derive information-theoretic limits of communication networks and suggest communication strategies that attain those limits. Using tools from information theory, this project further studied the fundamental tradeoff between performance, robustness against nonlinearities in the devices, and implementation complexity, aiming at novel encoding and decoding algorithms that can be implemented in hardware.

The project's research outcome can be divided into five lines of research, two concerning fundamental limits of wireless communication systems, and three concerning the optimal design of mobile devices:

1. Performance limits for noncoherent wireless networks

Most theoretical analyses of wireless communication systems assume that a genie provides the communicating devices with perfect knowledge of the channel propagation characteristics. This assumption, typically referred to as "perfect channel-state information", is overly-optimistic since it ignores the overhead incurred by the channel estimation. In our work, we accounted for the training overhead incurred by estimating the channel. Firstly, we developed a novel lower bound on the channel capacity of point-to-point wireless communication channels when the devices have imperfect channel-state information. Secondly, we explored how the lack of channel-state information affects the transmission rates achievable in wireless networks. Our results demonstrate that in dense wireless networks, i.e. in networks where the number of interacting nodes is large, the channel capacity is bounded in the SNR. Thus, communicating in such networks is highly power inefficient. This demonstrates the limitations of dense wireless networks and accentuates the importance of an accurate channel estimation.

2. Performance limits of short-packet wireless communication systems

We developed rigorous approximations of the maximum transmission rate at which packets of a given length can be transmitted with a decoding error probability not exceeding a given value. The approximations are accurate already for packet lengths of a couple of hundred symbols and are therefore relevant for short-packet wireless communication systems, i.e. communication systems that exchange short packets. The transmission of short packets is highly relevant for next-generation wireless systems. Indeed, the focus of future wireless systems will not only be on high data rates, but also on the introduction of new wireless modes, such as ultra-reliable communication and massive machine-to-machine communications. Applications for such communication modes include critical connections for industrial automation or reliable wireless coordination among vehicles. All these applications have in common that they have stringent delay and reliability constraints, i.e. they require the transmission of short packets with high reliability. Most information-theoretic analyses of wireless communication systems are based on channel capacity or outage capacity, which are defined asymptotically in the limit as the packet length tends to infinity. Such analyses provide accurate performance estimates when long error-correcting codes can be employed, but they may be very inaccurate for short-packet wireless communications. Furthermore, transmission strategies that perform well with respect to channel capacity or outage capacity may actually perform poorly when only short error-correcting codes are permitted. Our analysis of the maximum transmission rate does therefore not only yield more accurate performance benchmarks, but it also provides system designers with better guidelines on what transmission schemes are optimal to transmit short packets.

3. Practical signal constellations

We studied the rate loss induced by using a finite signal constellation instead of capacity-achieving Gaussian codebooks. We further studied the rate loss when the communication system uses Bit-Interleaved Coded Modulation (BICM) to transmit its data. BICM is used in many of the current wireless communications standards and is thus highly relevant for practical systems. Among other things, we proved that for BICM a so-called Gray labeling is optimal at high SNR.

4. Finite-precision decoders

Nowadays, most receivers employ digital signal processing techniques. Consequently, the analog received signal must be first sampled and quantized using an analog-to-digital converter (ADC). In this project, we have studied the limitations of ADCs and their impact on the transmission rates. To this end, we have analyzed the performance of practical scalar quantizers and compared it with that of optimal vector quantizers, whose complexity may be prohibitive for mobile devices. Our results show that uniform scalar quantizers may be a good option for mobile devices, especially if sufficiently high resolutions are feasible. They further show that dither should be avoided when communicating at small SNR.

5. Error correcting codes

Generalized Low-Density Parity Check (GLDPC) codes are a powerful class of error-correcting codes with favorable properties regarding minimum distance, error floor, and behavior for short blocklengths. They are therefore a promising candidate for mobile devices. Unfortunately, the analysis of GLDCP codes is a challenge, which is why often suboptimal decoding strategies are considered that are easier to analyze. However, these decoding strategies incur significant performance losses. In our work, we proposed an algorithm, named Probabilistic Peeling Decoder, that accurately and efficiently estimates the performance of GLDPC codes. It is therefore a useful tool for the analysis of GLDPC codes that will help coding theorists to optimally design error-correcting codes for next-generation wireless systems.

The obtained results provide a theoretical understanding of how future wireless communication systems and mobile devices should be designed optimally. They further provide system engineers with performance benchmarks against which practical systems can be tested. Thus, our results support system designers in the design of wireless communication systems that optimally trade off between spectral and energy efficiency. From a socio-economic perspective, MobileNET directly contributes to a more efficient design of future wireless communication systems that can cope with the demands of the Internet of Things. The socio-economic value of the Internet of Things is probably uncontested. For example, by connecting not only mobile device users but also objects, wireless communication systems will increase traffic safety by allowing vehicles to communicate with each other, improve medical services at a reduced cost by enabling remote medical treatment, or facilitate industrial automation. This will not only benefit society, but it will also offer a multitude of business opportunities.