International Technology Alliance

Responsibilities of Prof. Leung’s Group at Electrical Engineering, Imperial College

2011-2015

In this phase, Prof. Leung's group focuses on the fundamental research for network performance improvement in the context of coalition interoperable secure and hybrid networks. Tactical networks are designed to have the functionality to support mission needs, but they may not always be appropriate or adequate to form the sole provider of services to the military, with civil and commercial infrastructure services being brought into use to operate in conjunction with the military systems. Hence, research on hybrid networks that exploit and interoperate with commercial wireless networks is critical, in particular to develop the mathematical abstractions, models, and frameworks that will enable adaptive control of the behaviour of hybrid wireless networks for both unicast and multicast applications.

Collaborating with colleagues from IBM, UMass, ARL, in this phase, Prof. Leung's group leads two research topics: (i) Tomography of Hybrid Coalition Networks; and (ii) Mobile Micro-cloud, with the objective to support services and optimize network resources that are vital to coalition forces in the field.

1. Tomography of Hybrid Coalition Networks (Prj 1.1, TA5)

Hybrid wireless networks must be reliable, provide good performance, and be robust to disruptions. In order to provide these qualities, it is essential to have accurate and timely knowledge of the state of the network. Moreover, when anomalous behavior is detected, it is crucial to diagnose and locate the source of the anomaly. In the case of service disruption due to failure or lost connectivity of certain physical nodes, knowledge of where the failures reside will enable fast service recovery by migrating the affected services and/or routing around the affected region. However, direct performance measurement of internal network elements is not always feasible due to traffic overhead and lack of support at internal elements. This issue is aggravated in a coalition setting by security policies that block direct access to internal elements of other domains. This poses the challenge of how to design light-weight monitoring techniques that can reveal internal network state with minimum internal network support.

We propose to address this problem using network tomography, which refers to the methodology of inferring internal network characteristics from external, possibly aggregate measurements. Most Existing results on network tomography assumes that a large number of network elements are equipped with the monitoring functionality, which causes massive monitor deployment cost and traffic load. Emphasizing that the monitor deployment budget is limited or only certain nodes can be selected as monitors, we address the following closely-related sub-topics.

(1) Considering real routing constraints that each measurement path must be cycle-free, what are the necessary and sufficient network topological conditions for identifying all links metrics in a network with 2 or more monitors?

(2) For a fully identifiable network with given monitor deployment, how should we construct end-to-end measurement paths and solve each individual link metric efficiently?

(3) Given the budget of K monitors (the network is not fully identifiable no matter how to place these K monitors), what is the strategy to deploy these limited monitors such that the number of identifiable links is maximized?

(4) When the links in a network are characterized as random variables, how should the path measurements be performed, and what is the measurement redundancy that should be provided to increase the estimation accuracy?

(5) When the topology of a network changes overtime, how should we deploy the monitors such that the average number of identifiable links are maximized in the time domain?

This research topic belongs to Project 1 (Hybrid Networks: Performance and Metrics) within TA5 (Coalition Interoperable Secure and Hybrid Networks).

2. Mobile Micro-cloud (Prj 2.1, TA5)

In this project, we aim to develop cloud technologies to manage data services in tactical networks, in order to improve small unit operations by providing critical, timely, and mission relevant Situational Awareness. Different from conventional cloud technologies, we need to consider the security requirements of different forces in the military network.Additionally, some processors may be placed in an ad hoc basis and are interconnected with wireless connections. We therefore also need to study the optimal placement of different applications in the cloud and the scheduling of their workloads, both with dynamically changing link and node capacities. We specifically address the following sub-topics:

(1) Security-aware application partitioning and mapping. The security requirements of different forces impose additional constraints in formulating the application placement problem. This part of work intends to study how different security policies translate to such constraints, and how to solve the application placement problem subject to such constraints.

(2) Application placement and workload scheduling with incomplete network and processor capability information. This part of work first focuses on developing an appropriate model to take into account the randomness and uncertainty of link and node capacities. Following that, optimization methods of application placement and workload scheduling with incomplete information are studied.

This research topic belongs to Project 2 (Security/Network Management and Control) within TA5 (Coalition Interoperable Secure and Hybrid Networks).

2006-2010

In the first phase, Prof. Leung’s group mainly evolved in Task 1 of Project 1 in Technical Area 1. This task was to study the performance and design wireless ad-hoc networks using advance antenna techniques. The multiple-input multiple-output (MIMO) technique, which provides spatial diversity and multiplexing gain in wireless networks, can be applied to combat interference mitigation and reduction in interference-limited environments. Therefore, to meet the needs for future tactical communications, it is advantageous to equip the associated wireless ad-hoc networks with MIMO antenna capabilities.

The main objective of this task was to study the impacts of the physical-layer technique, e.g., MIMO, on the design of upper-layer protocols, including medium access control, packet scheduling, power control, routing, transport protocol and ultimately the overall quality of service in a wireless network.

1. Distributed Beamforming/Power Control for Relay Ad-Hoc Networks

Motivation of relay ad-hoc networks:
- Spatial diversity is desirable to combat fading
- Relay ad-hoc network, or cooperative network, yields spatial diversity without requiring to implement multiple antennas at each node

Key question of relay ad-hoc networks:
- How can each relay node decide its weighting factor in order to maximize the receive signal-to-noise ratio in a distributed way

Issues for Investigation:
- Design an approach of the distributed beamforming technique
- Develop performance analysis for the proposed beamformer by using information-theoretic measures.

2. Impact of Cooperative Transmission (CT) on Routing Decision

Issues for Investigation:
- Quality improvement of wireless links by using CT
- The distributed strategy of relaying node selection
- The effect of improved link quality to the routing decision

Furthermore, our research work on Network Monitoring and Troubleshooting in MANETs evolved from Task 1 of Project 2 at Technical Area 1.

3. Monitoring and Troubleshooting in MANETs

The ability to monitor network components and troubleshoot is crucial to the reliable and robust operation of military mobile ad hoc wireless networks. Our main objective was to allow network operators to quickly identify sub-optimal or abnormal behaviors of a network, and to correct the problem without requiring too much manual efforts.

The reason why it is difficult to analyze a performance problem in a multi-hop wireless network is that there may be multiple potential causes that interplay to produce observable symptoms. For example, consider a simple case when mobile nodes in a MANET cannot be reached. There can be many reasons for this symptom - the node's NIC may not function properly; the node may have moved out of range; there may be severe wireless link interference; or other intermediate nodes may be at fault. Furthermore, a problem in military networks should be understood in the operation environment. For example, in a densely deployed sensor networks, the failure of a single sensor node may not be critical, whereas in a tactical MANET, maintaining connectivity is of critical concern. The goal of this task is to develop efficient and scalable network monitoring/troubleshooting mechanism to enable the following:

- Fault localization and wireless root cause analysis
- Cooperative local monitoring of selfish behaviors and misconfigurations
- Proactive discovery of critical components that can be enforced to improve the overall performance of the network.
- Analysis of impact of network failure on quality of information

4. Optimal resource allocation for mission-oriented sensor networks using network utility maximization (TA3)

Optimal resource allocation for mission-oriented sensor networks using network utility maximization. The network utility maximization (NUM) is a very useful tool to identify the optimal parameters with varying resources and constraints. The main challenge in solving the NUM-related problems is the proof of the convexity and the guarantee of the global optimum of the multi-dimension coupled optimization, which has massive contributions to the topic of distributed optimization algorithms in general with specific applications to efficient resource allocation and configuration in mission-oriented sensor networks in particular.

We investigated the integration of NUM techniques with power control and sensor assignment to achieve a realistic framework for sensor network configuration. This was one of the first NUM-related projects that jointly considers multiple degrees of freedom. We also moved beyond the NUM framework in circumstances which dictate different solutions. Resolving these issues brings the techniques a step closer to practical use in military environments. We emphasize that the proposed work is in the context of sensor networks, however, the underlying techniques, principles and theoretical foundations studied can be generally applied to wireless ad-hoc and mesh networks. To summarize, the NUM related research topics include:

- Optimal air-time allocation
- Optimal scheduling with power and rate allocation
- Fundamental feasibility issue of decomposition of joint optimizations
- Integrate scheduling and bandwidth allocation with senor assignment and power control
- Data aggregation/compression with NUM