My long-term research goal is to effectively construct intelligent systems for large-scale, open and dynamic environments. I particularly focus on how to endow individual autonomous agents with the ability to act and interact in flexible ways and with effectively engineering systems that contain both humans and agents. I am interested in these issues from their fundamental principles, through design, to their application in particular problem domains. Specifically, I am interested in developing techniques for, and models of, automated bargaining, coalitions, trust and reputation, human-agent collectives, coordinated action and crowd sourcing.
I am passionate about pulling this work through into real-world applications. I was the lead system architect for one of the world's first industrial deployments of multi-agent technology, in the area of electricity transportation, and have subsequently been involved with deployments of agent-based solutions for: environmental sensor network monitoring in the Briksdalsbreen glacier in Norway and the River Solent in Southampton; controlling engine manufacturing lines at Daimler-Chrysler; workforce scheduling within BT; interpretation of seismic data for oil exploration with BHP Billiton; scheduling the repair of aircraft engines with Rolls-Royce; managing supply chains for defence applications, situational awareness and sensor management for unmanned aerial vehicles with BAE Systems; and wind prediction for the British Olympic sailing team. All of these systems have been used operationally and provide new capabilities or greater efficiencies to the organisation deploying them.
I helped create two start-ups. At Aerogility, formerly Lostwax, I am involved with the design and roll-out of their decision support tool suite that optimises the scheduling and maintenance activities for fleets of aircraft and allows forward simulation of strategic policy and resourcing choices (used by: BAE Systems for the Tornado, Harrier, Typhoon, Nimrod and Hawk fleets; Lockheed Martin for the F-22 Raptor and F-35 fleets; Sikorsky Aerospace Services for the S-92 fleet; and Boeing for the TLCS UK Chinook fleet). At variab.ly, I work on the design of negotiating algorithms for e-commerce marketplaces.
2015-2016: Head of Department, Electronics and Computer Science, University of Southampton.
2014-2016: Regius Professor of Computer Science, Electronics and Computer Science, University of Southampton.
2010 - 2015: Chief Scientific Advisor (CSA) for National Security to the UK Government
This involved working with a broad range of departments and agencies both in the UK and abroad. Responsible for scientists and engineers who work in the natural, physical, information and social sciences. Oversaw a research portfolio that delivered new operational capabilities; employed new mechanisms for delivering innovative technologies in partnership with range of companies (both large and small), academia and government labs; and championed science at board level. Inaugural post holder. Work closely with Sir Mark Walport, and previously Sir John Beddington, as a member of the network of departmental CSAs that provide scientific advice to government.
1999 - 2014: Professor of Computer Science, Electronics and Computer Science, University of Southampton.
2008 - 2010: Associate Dean (Research & Enterprise), Faculty of Engineering, Science and Maths, University of Southampton.
2001 - 2008: Deputy Head of School (Research), Electronics and Computer Science, University of Southampton.
1998 - 1999: Professor of Intelligent Systems, Electronic Engineering, Queen Mary, University of London.
1995 - 1998: Reader, Electronic Engineering, Queen Mary, University of London.
1989 - 1995: Lecturer, Electronic Engineering, Queen Mary, University of London.
1988 - 1989: Research Assistant, Electronic Engineering, Queen Mary, University of London.
Access to communication technologies has reached unprecedented levels and has made computational systems an interwoven feature of everyday life across the globe. Computational devices monitor our health, entertain us, guide us and keep us safe and secure. We are now massively interconnected via digital means, we routinely rely on computational devices and we increasingly invoke services residing in a global cloud infrastructure. However, this explosive growth in connectivity, devices and online services is only a precursor to an "era of ubiquity", where each of us will become increasingly dependent upon a plethora of smart and proactive computers that we carry with us, access at home and at work, and that are embedded into the world around us. As computation increasingly pervades the world around us, it will profoundly change the ways in which we work with computers. Rather than issuing instructions to passive machines, we will increasingly work in partnership with highly inter-connected computational components (aka agents) that are able to act autonomously and intelligently. Specifically, humans and software agents will continually and flexibly establish a range of collaborative relationships with one another, forming human-agent collectives (HACs) to meet their individual and collective goals. This project is funded by EPSRC and involves the Universities of Oxford and Nottingham, as well as BAE Systems, Rescue Global, PRI Limited and the Australian Centre for Field Robotics.
This project aims to develop control and planning algorithms for future applications of autonomous vehicles. In particular, it will investigate solutions to problems that require multiple robotic platforms to act together, in a coordinated fashion, so that the best use is made of their combined resources to tackle the task in hand.& To achieve to this, the project is adopting a practical multidisciplinary approach, in which multiagent planning and coordination algorithms will be developed and deployed on real robotic platforms, including Autonomous Ground Vehicles (AGVs) and Unmanned Aerial Vehicles (UAVs). These platforms will operate under the principle of flexible autonomy, in which robotic platforms will operate in a fully autonomous manner when appropriate, while still being guided by human involvement when key operating decisions need to be made.
This project will create a shared, facilitated learning environment in which social scientists, engineers, industrialists, policy makers and other stakeholders can research and learn together to understand how better to exploit the technical and market opportunities that emerge from the increased interdependence of infrastructure systems. The Centre will focus on the development and implementation of innovative business models and aims to support UK firms wishing to exploit them in international markets. The Centre will undertake a wide range of research activities on infrastructure interdependencies with users, which will allow problems to be discovered and addressed earlier and at lower cost. Because infrastructure innovations alter the social distribution of risks and rewards, the public needs to be involved in decision making to ensure business models and forms of regulation are socially robust. As a consequence, the Centre has a major focus on using its research to catalyse a broader national debate about the future of the UK's infrastructure, and how it might contribute towards a more sustainable, economically vibrant, and fair society.
This project developed intelligent agents (and other machine learning techniques) within the smart grid in order to reduce energy use within domestic settings. The project brought together an interdisciplinary team comprising experts in the fields of intelligent agents and multi-agent systems (School of Electronics and Computer Science), renewable energy and energy efficiency in the built environment, and human factors in the design of automated control and feedback systems (Sustainable Energy Research Group and Transportation Research Group in the School of Civil Engineering and the Environment) at the University of Southampton. The project was funded under the EPSRC "Transforming Energy Demand through Digital Innovation" call.
This project explored the issues associated with the decentralised control, operation and management of future generation electricity networks. It was targetted at scenarios in which micro-generation and storage capabilities are ubiquitous, where intelligent sensing devices allow users to make informed choices about the control of devices in their home, and where producers and consumers are connected via a series of dynamically negotiated supply contracts. This was an industrially funded project from a Hampshire-based company.
This project developed techniques, methods and architectures for modelling, designing and building decentralised systems that can bring together information from a variety of heterogeneous sources in order to take informed actions. To do this, the project took a total systems view on information and knowledge fusion and considered the feedback that exists between sensing, decision making and acting in such systems. Moreover, it achieved these objectives in environments in which: control is distributed; uncertainty, ambiguity, imprecision and bias are endemic; multiple stakeholders with different aims and objectives are present; and resources are limited and continually vary during the system's operation. This work was targetted at the domain of disaster management. ALADDIN was a multi-million pound multi-disciplinary research project funded by a BAE Systems and EPSRC strategic partnership. It involved a number of leading research groups from Imperial College London, University of Southampton, University of Bristol, and Oxford University.
This project addressed the challenge of developing effective and computational efficient inference and coordination algorithms in order to allow multiple mobile (and stationary) sensors to form agile teams such that they can efficiently represent, explore and search challenging, uncertain and dynamic environments. Working in collaboration with the University of Oxford, the project combined fundamental theory, algorithms and methodologies from the fields of multi-agent systems, decentralised control and Bayesian inference to allow physically distributed autonomous sensors to make effective, timely and coordinated decisions. This project was funded by the SEAS DTC.
Markets are dynamic places, in which participants trade their goods and services with an intention to maximise their own utilities. Now through such trading, highly efficient resource allocations can be attained in dynamic and uncertain environments. Given this, this project applied market-based paradigms to the design, control and evolution of complex distributed computational systems. The targetted applications included resource allocation in utility data centres, decentralised control of content delivery and multiple robotic systems. Funded by EPSRC (under the Novel Computation Initiative).
This project addressed the need for an integrated approach that allows such sensor networks to manage their energy requirements, and where possible, harvest energy from the local environment, whilst simultaneously coordinating their activities in order to achieve system-wide aggregate goals. Funded by the Data and Information Fusion Defence Technology Centre.
This project was concerned with investigating and developing the basic mechanisms that enable collectives of software agents to self-organise, self-repair and self-optimise in response to dynamic environments. Such autonomic behaviour is needed to ensure the agents are able to best achieve both their individual objectives and the objectives of the collective as a whole in the face of a constantly changing and highly uncertain operational environment. In such cases, the resources available (communication and computation) to the agents are in a constant state of flux and the set of agents in the system is constantly changing. In such cases, it is impossible for the a priori system design to continue to be maximally effective because many of its operational assumptions and parameters are changing. Thus, the system can gradually degrade its performance or it can endeavour to respond to such changes by reorganizing itself in order to best achieve its objectives. Funded by the Data and Information Fusion Defence Technology Centre.
This project developed techniques and methods that enable the automatic establishment, maintenance and operation of supply chains in highly dynamic, multi-stakeholder environments. It also focused on the associated supply chain business models for such agile and dynamic environments. In more detail, the various actors within the system, each with their own aims and objectives, were represented as autonomous software agents that interact in a number of on-line markets in order to procure the goods and services they require in a timely fashion. The markets were also represented as autonomous agents and so adapt their offerings and their terms and conditions in order to attract traders and better differentiate themselves from similar competing markets. Given this, the ensuing supply chains will need to be autonomic --- self-organizing, self-healing, and self-optimizing --- Â in order to cope with the high degrees of dynamism and uncertainty that are present in the system. Moreover, through its continual adaptation in response to change, the resulting computational economy will offer significant advantages to all its participants in terms of agility, lead-times, and profitability. To provide a specific illustration of this vision, this feasibility study examined the supply chain associated with engine aircraft repair and overhaul in conjunction with end-users at Rolls-Royce.
The aim of this project is to perform the formative research required to construct a reactive decentralised data fusion system and to demonstrate its value in industrially relevant applications. Such a system must fuse information from disparate and varied sources, whilst coping with unreliable data and limited communication bandwidth, in a time critical environment. The core of the project concerns the integration of agent technologies and Bayesian statistical methods. We are specifically interested in designing the mechanisms by which self interested agents will trade for information, computation resources and bandwidth. The goal of this work is to ensure that the self-interested actions of the individual agents results in desirable system-wide properties. The project involves three industrial partners, each of whom is providing a distinct application area: BAE SYSTEMS, Rolls-Royce and QinetiQ. The project is a Defence Aerospace and Research Partnerships (DARP) project, with joint funding from the Department of Trade and Industry (DTI), the Ministry of Defence (MoD) and the Engineering and Physical Sciences Research Council (EPSRC).
This project carried out the fundamental computer science research that is necessary to support the entire virtual organisation (VO) lifecycle as it exists for e-Science and the Grid. Specifically, this involved research into how services can be composed in order to achieve a particular objective; how to recruit an appropriate set of agents to meet service specifications; how to assign individual duties to the constituent agents; how to monitor the performance of the VO; and how to reconfigure the VO in face of changing circumstances. Funded by EPSRC (under the e-science fundamental computer science program) and undertaken in collaboration with Mike Wooldridge.
Within a ubiquitous environment, market-based approaches can be used to select the most appropriate material for a public display, depending on factors such as the audience'ÂÂs preferences and diversity of interest. Likewise, strategies used by agents to compete for customer attention should strive to be rational, based on contextual observations of user-preferences within the local environment, and should include a reward mechanism based on audience responses. Ubiquitous devices such as bluetooth-enabled mobile phones, can be used to uniquely identify and detect the presence of individuals within a localised environment, without the need for deploying bespoke hardware. BluScreen is an auction-based framework for presenting consumer advertisements is described, whereby agents (representing consumer advertisements) can compete for consumer attention, where consumer interest is determined through observations of ambient bluetooth activity.
This research developed techniques and algorithms for coalition formation in virtual enterprises (with particular emphasis on the Grid). Our work focused on the mechanisms by which coalitions can be formed, maintained and disbanded when they are no longer effective and on the impact of trust and reputation on such processes Funded by Welsh E-Science Centre, DTI and BTexact.
This project was concerned with developing flexible and robust methods for managing decentralised systems in general, and supply chain management systems in particular. We developed an agent-based control systems that: (i) operates in a decentralised way while exhibiting desirable system-wide characteristics; (ii) produces effective local decisions that contribute towards an effective overall system (emerging behaviour); and (iii) operates in a computationally efficient manner. Funded by the Data and Information Fusion Defence Technology Centre.
This project investigated the role of argumentation-based negotiation in a social context. It developed a complete formal model of an agent that is capable of using its social context to argue effectively. The application domain is one in which agents try to persuade one another to view information that they deem relevant. Funded by EPSRC.
The research investigated the role of intelligent agents in future generation, mobile communication environments. In particular, issues related to how agents can flexibly adapt their behaviour and their interactions to the characteristics of their current communication environment were explored. Funded by the UK's Virtual Centre of Excellence in Mobile and Personal Communications.
This project explored the use of agent technology for providing various forms of network service. In particular, investigated the use of both cooperative and competitive (market-based) approaches to tackling this problem. Funded by EPSRC (under the programmable networks program).
This project aims to design and build the infrastructure that makes customised informaton available to intermittently connected users. For that purpose, we will investigate the use of agents as autonomous intermediaries between nomadic users and fixed infrastructure services. Application specific mobile agents, spawned from users'ÂÂ PDAs, will migrate to the infrastructure in order to autonomously undertake their task. These agents will be used to provide users with the means to access and exchange information, in an ad-hoc and secure manner, while on the move. Multi-agent interaction protocols, such as negotiation and cooperation, will help preserve the security of the environment. Open hypermedia techniques, and in particular link services, will be investigated in order to deal with information management in this context; in particular, these techniques will be used to filter and present information according to the users'ÂÂ needs. Funded by EPSRC (under the DIM program) and DERA.
This project is concerned with the process of managing intrusiveness in pervasive computing environments. In particular, we intend to investigate the use of argumentation-based negotiation as mechanism for managing intrusivess in a context sensitive fashion. Funded by the IST program under the Disappearing Computer initiative.
This research will investigate and develop techniques by which software agents can acquire sufficient knowledge to negotiate effectively on behalf of their user in a range of electronic commerce scenarios. The research will exploit and extend work in the Knowledge Acquisition community in order to determine exactly what knowledge an agent needs to be endowed with to negotiate on behalf of its user, what techniques are appropriate for capturing this knowledge, and how the agent's performance can be evaluated against its users' expectations. Funded by Hewlett Packard Research Labs.
Formal investigation of the use of commitments and conventions in cooperative problem solving. Described the cooperative process from detecting a need for cooperation, through team formation and planning, to execution. Show how such models can be used to design and build architectures for cooperating agents. Funded by DERA Malvern.
Investigated the role and function of brokers and other forms of intermediary in an information economy. Particular attention was paid to issues of dynamic self-organisation and scaleability. Funded by Marconi Communications.
This EPSRC funded collaborative project aims to apply game theoretic techniques to the design of negotiation algorithms for use by autonomous software agents in electronic commerce. The other partners are the University of Liverpool (Mike Wooldridge) and the ESRC Centre for Economic Learning and Social Evolution (Ken Binmore and Nir Vulkan).
This EPSRC funded project investigated a novel approach for managing Connection Admission Control (CAC) in Asynchronous Transfer Mode (ATM) telecommunication networks. The project addressed the application of an intelligent multi-agent system, in which agents were modelled as self interested decision makers in a computational market, to network level resource allocation.
Investigated the role of negotiating agents in telecommunications systems. In particular, it dealt with multi-attribute negotiations (some of which are competitive and some of which are cooperative) in a dynamic and uncertain environment. Funded by Nortel Networks
This EPSRC funded collaborative project was concerned with the design and implementation of a multi-agent system for managing the information and services (e.g. use of software tools) within a virtual research laboratory across the sites of three UK universities (QMW , Durham and Strathclyde).
This collaborative project was concerned with the use of multi-agent systems in the management of business processes. Concentration was given to the problems involved in coordinating semi-autonomous departments of a large organisation or multiple organisations by utilising an agent'ÂÂs autonomy and managing their dependencies through automated negotiation.
Used cooperating agents to extract interesting features from very large volumes of scientific data. Agents were conceptualised and implemented using the belief-desire-intention paradigm (under dMARS). Project undertaken in collaboration with BHP.
European collaborative project which built some of the first real world applications of multi-agent systems. Applications were in the broad area of process control and, in particular, electricity management and particle accelerator control
Research Student Supervision
Baker,C, Agents and disaster response
Beck,Z, Coordinating UAVs
Pruna,R, Game theory and machine learning
Truong,N, Incentive engineering for crowd sourcing applications
Zenonos,A, Incentives and crowd sourcing
Khan,M, Large scale coordination
Manino,E, Machine learning and crowd sourcing
Panapopoulos,A, Machine learning and energy systems
Augustin,A, Multi-agent coordination