Project Aims

---

This project aims to develop 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 needs to take a total systems view on information and knowledge fusion and to consider the feedback that exists between sensing, decision making and acting in such systems. Moreover, it must be able to achieve 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.

More specifically, the main aims of the project are:

• To devise techniques that enable an actor to effectively balance acting and information gathering in dynamic, uncertain, multi-actor environments.
• To devise techniques that enable an actor to fuse, in a decentralised manner, inter-related information that is uncertain, incomplete, imprecise and ambiguous.
• To develop machine learning algorithms that are efficient and effective in dynamic, multi-actor
  environments that are uncertain and incomplete.
• To develop coordination mechanisms that enable collectives to plan and act collaboratively in order to achieve common goals.
• To develop methods for modelling and predicting the system behaviour that will ensue from specifications of the local behaviour of the individual actors.
• To develop mechanisms that ensure desirable overall properties emerge based on local actions and views.
• To develop decentralised system architectures that can operate effectively in uncertain and dynamic environments and that are robust, scaleable and flexible in their operation.

To ensure the specific methods and techniques developed in the research fit together to give a coherent whole, the project will develop a number of software demonstrations.

Application Context: Disaster Management

The circumstances during and after a disaster require us to obtain as clear an assessment of the situation as is possible and then act on this assessment to alleviate short term suffering without compromising, if possible, longer-term goals to return the disaster stricken environs to ‘normality’. After any disaster there is a wide range of interested parties, public and private, anxious to do all within their power to recover the situation, but the objects of their attention and what they consider to be ‘helpful’ are not necessarily consistent. Those agencies and authorities, who are drawn from outside the disaster to help, will often focus on different aspects of the unfolding scenario, building their own partial understanding of ‘ground truth’ oriented towards their potential contribution and its effective timeline. For each stakeholder in the disaster recovery scenario their assessment of the situation will therefore be based on their pre-disaster knowledge and data gathered (not necessarily by themselves) post-disaster. Such data will be:

• Uncertain
• Changing / Dynamic
• Incomplete

In such situations, automated systems and decision aids are needed to support the overloaded human decisionmakers, within the varied agencies and authorities, deal with the flow of often imperfect information and focus on timely instigation of ameliorative actions. To exploit what we have available it will be necessary to link the planning and decision processes with the situation assessment. This would provide the facility for:

• Action to be based on the latest situation assessment or prediction.
• The identification of actions that could provide better situation assessment (search for info)
• Development of strategies that avoid regions where information is weak or unknown (the safe option)
• Through communication the co-ordination of components to provide
  1. better coverage,
  2. better use, and
  3. faster response given limited resources.

Given this context, the kind of problems we are concerned with are characterised by:

• A wide range of data sources: The data that will be used to build up the assessment will come from a multiplicity of sources and be of diverse formats (unstructured text through to video images) as well as of uncertain accuracy and objectivity. Baseline pre-disaster data may not always be available.
• The assessment is open-ended: The complications that arise in disaster scenarios are always specific and until recognised the generic response cannot be instantiated (e.g. the risk of ensuing health problems is recognised, the particular form it may take in any particular disaster is not so readily predictable).
• Time is a key factor: This is both in terms of timeliness of results and in terms of the dynamics of the evolving situation.
• The outputs are diverse: Ranging from the identification of specific problems with clear response to the recognition of trends that require the underlying issue to be established before a response could be organised.

This is against a back-drop where:

• Many events are occurring simultaneously with differing priorities for attention and need for response.
• The situation is continuously evolving and hence associated priorities and decisions need to be revisited
• We are operating in an uncooperative environment that may as a result of the disaster be disrupting communications and transportation networks and presenting changing logistical challenges.

In terms of the ensuing DDIS, it needs to be capable of delivering the following functionality:

• Co-ordination of different assets under incomplete information.
• Distributed decision-making and control between assets: Decentralisation of the control process
  provides robustness and degrees of autonomy, allowing a local agent to independently respond to local stimuli without the need for central authority or control (thus increasing flexibility of response).
• Decentralised information fusion: We will need the ability to fuse all available information to produce a picture that is more complete, more accurate, more timely (given dynamic data) and share that picture across all stakeholders. Decentralisation of information allows greater flexibility, robustness and scaleability and hence the ability to support a wider range of disaster scenarios