In the army domain, a cold war type scenario was rejected in favour of a peace enforcement scenario. Support is focussed at the divisional level in a multi-national force intervening between two bordering states. AI tools are provided to assist with manoeuvre and fire support planning.
Generally speaking, the user supported is the Division Commander and his staff. More specifically the primary user envisaged in the G2/G3 merged and the second user is G3 plans.
The objective of Automated Report Analysis (ARA) is to reduce the volume of data that needs to be assimilated by the command team and to ensure that important information is given priority.
The ARA Automatic Message Processor (AMP) takes simulated reports for the land or naval domain generally consistent with the ACP127 and AdatP3 standards. A set of mini-parsers have been developed using Yacc++ to automatically extract content from structured messages as the basis for compiling the Wide Area Picture (WAP). The current context determines the parsing strategy via a knowledge-based controller which triggers an appropriate set of mini-parsers. The AMP Agent plays the role of Intelligence Officer, processing incoming reports, providing extracted information to the WAP Agent and handling house-keeping of reports.
The WAP is represented in the ObjectStore object-oriented database and this is underpinned by the GRACE Common Model (GCM), an agreed object model representing the information of interest in the command and control problem domain and loosely based on the ATCCIS model. The WAP agent is responsible for maintaining the definitive Wide Area Picture and distributing information to Planning Agents in accordance with their subscription requests.
The following support facilities for army DSPT have been demonstrated:
Calculation of mobility corridors
Course of Action comparison
Manoeuvre planning
Close air support planning.
Digital terrain data in DIGEST format is stored in an Illustra database (an RDBMS extended with a spatial ‘datablade’). This data can be displayed as a map with switchable terrain feature overlays. The terrain features are used to identify go and no-go areas. From this information the Terrain Analyser Agent identifies mobility corridors of a defined width, necessary for a unit of given size, are constructed automatically using a geometric technique (the Generalised Voronoi Diagram). This identifies possible routes and nodes between routes, which provide a starting point for Course of Action construction, which is necessarily a co-operative process between the machine and the human commander, using his experience and creativity.
Support is provided for comparing own and enemy Courses of Action using Weapon Effectiveness Indices and Weighted Unit Values. This combines subjective assessment of relative weapon effectiveness with probability analysis of outcomes. The COA Comparison Agent is a tool to support the command team to construct a COA, rather than trying to automate this process.
Construction of a detailed manoeuvre plan is supported by combining constraint satisfaction with real-time heuristic search. This enables complex temporal, spatial and tactical constraints and objectives to be considered. As part of this work the following tools have been developed:
a Fuzzy Knowledge Manager to deal with knowledge bases representing aspects like morale, visibility and the weather.
Temporal Knowledge Manager for representing and reasoning with time constraints
a FORCE Browser, for analysing and modifying the Task Organisation (hierarchy of military units)
Plan Viewer, presenting unit activities and interactions versus time
For planning Close Air Support, simulated annealing is used in an anytime formulation, to generate the best allocation of aircraft from different bases to different targets that is possible in the time available. Rigorous techniques have been developed for updating such a plan in the light of changing information.
 |
 |
 |
| Produced by: Peter Martin / Russell Gordon /
Mike Pockney Updated: 30 March 1999 Copyright Logica 1999 |
 |