Research programs

On a global scale, traffic accidents are one of the leading causes of serious injury or even death. In 2004, the World Health Organization estimated that 1.2 million people lost their lives in traffic accidents around the world, making it the ninth leading cause of death worldwide. It is projected that traffic accidents will grow to be the third leading cause of death in the world by 2020.

Recent advances in wireless communication technology, e.g. WiFi or Wireless LAN, have revolutionized personal mobile connectivity. In much the same way, these technologies are on the verge of revolutionizing wireless inter-vehicular communication. In particular, Dedicated Short Range Communications (DSRC) is emerging as the communications standard for future intelligent vehicle systems. These systems are currently being adopted around the world as a means to: Improve safety and save lives lost to traffic accidents; Reduce the economic impacts of traffic accidents and congestion; Reduce the environmental impact of congestion.

The Institute for Telecommunications Research (ITR) and its industry partner, Cohda Wireless, are pioneering state-of-the-art DSRC technology that uses the latest IEEE 802.11p standard to enable wireless vehicle-to-vehicle communications. With this technology, each vehicle is able to broadcast its position, velocity and direction information to all other vehicles within a certain range. Each vehicle is thus able to integrate this information with map data to build up a picture of the traffic environment, which can then be used to warn the driver of potential hazards (e.g. collisions, corners, traffic lights, give-way/stop signs etc.). This is however a very challenging networked environment, consisting of very many moving vehicles, with uncertain locations. The aim of the project is to develop robust approaches for ad-hoc vehicular networks, and integration of inter-vehicular communication and geographic information to provide novel hazard avoidance technology.

Space-time processing (using multiple antennas) is now recognized as a key to achieving reliable high data rate wireless communications and is being incorporated into the physical layer of many wireless standards. In this project we have introduced another degree of freedom: electromagnetic polarization.

At ITR, we are developing new algorithms and power-efficient hardware architectures for signaling, detection, demultiplexing, and decoding. In collaboration with the University of Melbourne, Princeton University and DSTO, these elements are incorporated into a cross-layer approach to utilizing polarization as a new resource for wireless access to broadband in the bush.

ITR has recently been involved in two gigabit-per-second wireless communications projects; one for single-carrier satellite downlinks and the other for short-range indoor networks using multi-carrier and multi-antenna techniques. Given increasing demand for broadband wireless communications, we are starting new research projects that address these areas, as well as continuing some of the existing work.

For example, free space optical communications, when combined with modern coding methods, can overcome scintillation effects caused by the atmosphere and achieve reliable communication at very high rates. One aspect of this involves the design of new receiver architectures capable of achieving slot timing and frame sync in PPM schemes, plus custom architectures for gigabit LDPC codec design.

Wireless data communications is becoming ubiquitous. To meet the demands of future high-speed wireless applications, systems are approaching fundamental physical limits, where implementation complexity is a major problem. Iterative information processing has emerged as the dominant low-complexity design paradigm.

The aim of this project is the optimization of data communications systems subject to constraints on computational complexity. The project builds on previous research accomplishments in these areas, aiming to formally include complexity constraints into the design of concatenated coding systems, and into the development of low-complexity iterative algorithms. The main research tasks are to formulate a mathematically tractable measure of complexity relevant for concatenated systems and corresponding iterative processing algorithms, and to incorporate this analytical complexity measure into the design of optimal, complexity-constrained code structures.

With higher data rates in wireless communication networks, new broadband services like DVD-quality video and CD-quality audio become relevant for wireless devices. Each application requires the network to deliver a specific Quality-of-Service (QoS) in terms of minimum errors and delay. Performance of broadband wireless communication networks is limited by availability of resources such as frequency bandwidth and transmission power.

A major network design challenge is to provide a wide range of QoS, given the limitations of wireless channels, and the limited available resources. Transmission schemes, adapting to instantaneous channel characteristics can significantly improve performance.

The main objective of this project is to increase the throughput of future wireless communications systems by adopting adaptive principles in the underlying communications protocols. The specific aims are to determine optimal adaptive transmission strategies for delay-limited block-fading channels, minimizing the outage probability; extract design guidelines for adaptive transmission schemes through a thorough theoretical understanding of the outage probability; and develop practical adaptive, low-complexity coding and decoding strategies that can perform arbitrarily close to the outage probability for delay-limited block-fading channels.

Network coding is a recent breakthrough, which uses coding rather than routing at the network layer for data transport. It has many advantages, including higher throughput, lower delay, and increased robustness This project will investigate the application of network coding to ad-hoc wireless communications networks. The goal is to use network coding, rather than complex ad-hoc routing protocols. It has already been realized by several researchers that the wireless channel itself can act as a kind of "free" network coding device.

Firstly, we will aim to find information-theoretic bounds on the throughput and reliability of wireless network coding. Secondly, we will aim to devise practical transmission schemes to exploit this possibility.Experimental work will be carried out using the SANLAB mobile ad-hoc networking laboratory. This project involves collaboration with researchers from Adelaide University and DSTO.

Network coding opens the door to many interesting possibilities for information security. The use of multiple transmission paths may increase robustness to denial of service or jamming attacks. It can also provide security against eavesdroppers.

The aim of this project will be to explore some of the security implications and advantages of network coding.

Network coding is particularly interesting for multicast scenarios, where many receivers require the same data from a single source. In this project, we will investigate the application of network coding to distribution of multimedia traffic. It is of interest to consider receivers with different capabilities that may require different resolution versions of the same source data. One example would be streaming a video source, where some receivers have high-resolution displays, and others are small mobile terminals.

The objective will be to develop rateless network codes supporting this kind of multi-resolution streaming data.

During the past few years ITR has developed a novel opportunistic rate adaptive medium access control protocol suited for ad hoc communication networks. The objective of this project is to develop "opportunistic routing" protocols to take advantage of the new MAC protocols. In this project we will attempt to identify the end-to-end "goodness of transmission conditions" along different paths between a given source and destinations. Once these have been estimated they will be compared to the requested QoS between a source and destination. The path that closely matches the QoS requirements can then be selected as the primary route between the source and destination.

Research problems that need to be addressed in this project include methods to estimate the "goodness of transmission conditions" between a given source and destination (this will be an extension of probability of good channel conditions in the current MAC protocol), effective collection and exchange of information required to create such measures and characterising QoS requirements of different nodes.

Advances in the development of special-purpose wireless ad hoc networks will provide opportunities for emerging wireless applications. Example applications with significant relevance to Australia include public safety and emergency-area communications. Cooperative communications is a new technology specifically targeting performance improvements in wireless ad hoc networks.

The aim of the project is the design, evaluation and implementation of efficient transmission strategies for reliable communication across cooperative wireless ad hoc networks. The project focuses on the innovative use of channel and network coding techniques to increase reliability, throughput, and coverage of cooperative wireless ad hoc networks.

The main task is to devise new transmission strategies for application across all nodes within a cooperative wireless network of arbitrary size and topology. The emphasis will be on efficiency, simplicity, and scalability to allow for practical implementation. Two key concepts are to employ multi-layer transmission schemes for cooperation and to exploit inherent properties of wireless channels for network coding.

Recent advances in mobile communication technology, efficient/portable power sources, and high speed computing have enabled the deployment of sensor and ad hoc networks, which are a collection of wireless terminals that can form a network in the absence of any fixed infrastructure support. Ad hoc networks have applications in defence and various emergency services. Recently, sensor networks have emerged as another interesting application of telecommunication networks, where a large number of sensors form a network to collect, process and distribute information.

However the theoretical and mathematical understanding of these networks is still in its infancy. Due to the lack of a solid theoretical understanding, the performance of current implementations of both ad hoc and sensor networks has fallen well below the generally expected performance. In large sensor networks, it is not realistic to envisage that a central entity would be capable of coordinating all tasks performed by a very large number of sensor nodes. Instead, they require distributed medium access and routing mechanisms to function effectively. Furthermore, there are strong dependencies between neighbouring nodes in sensor and ad hoc networks, while there are often no direct interactions between non-neighbouring nodes.

The lack of statistical tools to analyse large, distributed systems which exhibit strong spatial dependencies in a 2 dimensional domain is a major obstacle on the way to develop a clear theoretical framework to understand, design and implement sensor networks. Providing a comprehensive analysis toolset is thus a major aim of this proposal.The aims of our project are to:

1. Develop a theoretical framework, based on the theory of Markov Random Fields, to characterise the interactions between nodes of a sensor network in the 2 dimensional domain and define an energy function related to each of these interactions.
2. Use this theoretical framework to study the performance of sensor networks and the impact of selected network parameters (e.g. node density, power allocation, interference) or protocol mechanisms (e.g. routing,resource allocation) on the overall network performance. This will also lead to understanding inherent performance limitations of sensor networks.
3. Design distributed network protocol mechanisms, which optimise the network performance (e.g throughput, end-to-end delay, etc.) using this theoretical framework. We will address this more practical problem by taking into consideration requirements and issues specific to target tracking in defence or civilian applications, as well as biomedical applications of sensor networks.

In Free-Space Optical (FSO) communication, information data is transmitted wirelessly using laser beams. Such links have the potential to provide virtually unlimited bandwidth without the expense of buried fibre optic cabling. Unfortunately, FSO link reliability is highly dependent on atmospheric conditions such as turbulence, rain, cloud, dust and fog. These effects cause fading of the received laser beam. In typical FSO channels, a deep fade can cause the loss of millions of data bits. Improvement in reliability can be gained by using multiple lasers and multiple apertures to create a multiple-input multiple output (MIMO) FSO channel. However, the atmospheric effects in multiple-laser, multiple-aperture FSO systems are largely unknown, and experimental analysis is lacking.

A comprehensive characterisation of this channel will be crucial in order to exploit the full capabilities of MIMO FSO links. This project aims to investigate these aspects by collecting channel measurements from both terrestrial and (if available) satellite-to-ground FSO links equipped with multiple-lasers and multiple detectors. Key parameters of interest are the temporal and spatial fading and background noise statistics under a range of climatic and environmental variations, including daily and seasonal variations for both clear-sky and impaired situations due to dust or clouds, for a variety of propagation lengths. Using these channel measurements, statistical models will be developed that describe the spatial-temporal behaviour of MIMO FSO channels.

Although free-space optical (FSO) links can usually support huge data rates, their performance is affected by fadingdue to atmospheric turbulence, plus meteorological conditions such as dust, clouds and fog. Whilstfading due to atmospheric turbulence can be mitigated using multiple-lasers and multiple-apertures,the large signal attenuation due to fog and cloud cover poses a formidable challenge. It appears theonly way around these problems is to use an additional radio frequency (RF) link, creating a hybrid FSO/RF channel. Inparticular, high frequency Ka-band (18-40 GHz) or V-band (40-70 GHz) links are most favourable forthis purpose to maintain data rates comparable to that of the FSO link.

At these frequencies, the FSOand RF channels operate in a complementary manner to each other. For example Ka-band RF linksare less affected by fog or cloud, but can be seriously degraded by rain. FSO links do not penetratefog/cloud well, but are reasonably resilient to rain. Thus, practical FSO systems are likely to use ahybrid approach that intelligently combines the advantages of FSO and RF links.Despite the obvious intuitive advantages of hybrid FSO/RF links, little is known regarding how tomodel the combined statistical behaviour of the component channels, and even less is known on howto efficiently combine them to improve system reliability. The aim of project is to characterise the jointstatistics of hybrid FSO/RF channels under a variety of atmospheric conditions. This will involve collecting sample measurements from simultaneous FSO and Ka-band RF transmissionsover a long propagation distance (approximately 20km) and mathematical modelling of the observed data.

The use of MIMO and hybrid RF/FSO techniques has great potential in significantly improving the reliability of free-space optical communication.However, very little is known on how to efficiently transmit information reliably using these channels.

This project will involve an information theoretic analysis of these channels to determinekey design parameters that underpin efficiency and reliability of the transmission of information. In particular the project will investigate the affects of non-ideal photodetection, scintillation, spatial and temporal correlation on the fundamental limits of FSO communication. Novel transmission strategies will then be investigated that achieve as close as possible to these fundamental limits, whilst remaining feasible to implement in practice.

Computational neuroscience is the collective name for scientific research into understanding the brain that employs tools familiar to mathematicians, physicists, computer scientists and electronic engineers. Our brains operate very differently to digital electronic computers. One of the goals of computational neuroscience is to understand the physical mechanisms biology uses to acquire, process, communicate and compute information.

There are several levels that can be studied including the chemistry that occurs within a single neuron (brain cell), the overall electrical activity of a single neuron, and the behaviour of groups (populations) of neurons. Sensory neurons are neurons that acquire and process information about our environment and that allow our senses to work. One way to better understand our senses and neural coding is to model and analyse, using information theory, the features that are also present in artificial communication or sensor networks.

Some of the unsolved questions in this area include

1. Does unpredictable neural variability have an essential role in sensory neural coding, or is it like electronic white noise that needs to be filtered somehow?
2. Ingenious biological mechanisms, like the basilar membrane in the inner ear, nonlinearly transform incoming signals in a compressive way. Has evolution optimized these mechanisms in an information theoretic sense, given constraints like a need for energy efficiency in the cells that convert from stimulus to neural activity?
3. What is the best way to try to replicate natural neural signals in brain machine interfaces?

A Stochastic Pooling Network is a model that can be used to understand complex nonlinear interactions between random noise, redundancy and compression. The model is applicable in many situations, for example biological neural populations, nano-electronic circuits and distributed sensor networks.

There are many interesting possibilities for theoretical research in this area, for students with backgrounds in any of the following areas: mathematics (pure or applied), numerical computing, information theory, electronic engineering or physics.