EUCLID RTP 6.1 has produced an integrated demonstrator featuring two key innovations:
an agent-based architecture to integrate diverse artificial intelligence based applications, and
an integrated suite of command decision support tools applying AI technologies.
The integration architecture comprises:
CABLE for developing and running software structured as multiple co-operating intelligent agents. For time-critical applications, Real-Time CABLE Extensions enable agents to be developed that provide their services within a guaranteed response time.
FIONA to provide the user interface between one or more users and multiple agents, including a map-and-overlay display called the DOHP (Digital Overhead Projector).
Both CABLE and FIONA embrace the CORBA standard as implemented in Orbix, enabling multi-agent software to be distributed at run-time across a mixed network of Sun Solaris and PC Windows-NT workstations. CABLE, FIONA and the FUN Methodology for specifying and designing multi-agent systems have been used effectively across the GRACE consortium and are applicable to any problem domain characterised by distributed software, multiple users and an element of artificial intelligence. They are particularly well suited to intelligent decision support and modelling and simulation applications, whether military or not. At the time of writing 10 new projects, most in the UK, have exploited this architecture, with a total value of 2 MECU (£1.4M).
Some 14 tools are provided to support different aspects of army and naval situation assessment and planning, using a wide range of AI techniques, each suited to the problem at hand:
report analysis and situation assessment
AMP - Automatic Message Processing, using object-oriented parsing techniques to extract message content
WAP & WAPC - Wide Area Picture & Compilation, using fuzzy logic and clustering algorithms to identify significant enemy behaviours and groupings, and a publish-and-subscribe mechanism to notify other tools of changes to the WAP
GCM - GRACE Common Model; an object model used to represent the Wide Area Picture and encyclopedic data, so that all agents can exchange information in an agreed form
ADVERTISER - enabling the user to define rules to trigger alerts on significant changes in the situation
TTA - Tactical Threat Analysis, using rule-based techniques to represent possible enemy plans and what would be observed at each stage of those plans, to provide analysis of enemy activity
army decision support, planning and tasking
MADRID - Map Display; to store, display and manipulate vector feature data
TEA - Terrain analysis & mobility corridor construction; using a geometric algorithm to identify mobility corridors suitable for units needing a particular doctrinal frontage
COAC - Course of Action comparison, for comparing own versus enemy courses of action using Weapon Effectiveness Indices and Weighted Unit Values
MFSP - Manoeuvre Planner; this integrates a number of planning tools including scheduling using TIM, a synchronisation matrix display and simulation of the possible evolution of own and enemy activities, taking account of doctrine and the terrain.
TIM - Time manager; a software tool for representing and reasoning with time constraints.
FORCE - ORBAT Browser, for multiple users to view and change an ORBAT or Task Organisation.
RACAS - Resource Allocator; uses simulated annealing in an anytime formulation to allocate own aircraft and weapons to attack enemy targets.
naval decision support, planning and tasking
EC - Engagement Co-ordination; uses spatial reasoning, probabilistic reasoning and constraint-based techniques to plan a naval engagement
TeART - Terrain analysis; anytime analysis of: routes, missile range areas and intervisibility, in which solution quality is traded against timeliness.
TED - Terrain Exploitation and Display; a tool for automating the production of colour maps based on terrain elevation data.
MPB/SP - Manoeuvre Plan Browser and Situation Predictor; uses user-defined hypotheses to predict future positions of friend or foe units
MCFP - Manoeuvre co-ordinator and formation planner, using constraint-based techniques for reasoning about co-ordinated manoeuvres.
The tools were generally intended to support the human command team, by automating only those aspects of a task that are better suited to the machine. Each of the tools was evaluated and demonstrated using simulated data from a naval landing force scenario or an army peace enforcement scenario. The benefits typically found from using AI techniques compared with manual planning were:
automatic alerting to significant events or changes in the situation
quicker planning
consideration of more alternative plans
more constraints (time, space, terrain, resources) were taken into account.