Loading...
Please wait, while we are loading the content...
Similar Documents
Using artificial intelligence algorithms for high level tactical wargames and new approaches to wargame simulation
| Content Provider | Semantic Scholar |
|---|---|
| Author | Lucek, Stephen Collander-Brown, Simon |
| Copyright Year | 2018 |
| Abstract | Received 06 October 2016 Accepted 22 August 2017 Abstract—The Mission Planner is a decision-making toolset developed by NSC, a UK based provider of cutting‐edge training, modelling and simulation and consultancy services, for Dstl (Defence Science and Technology Laboratory, an executive agency, sponsored by the UK Ministry of Defence) currently applied at the tactical level of combat. It aims to support Dstl high intensity warfighting simulations by reducing or eliminating the need for complex pre-scripting of simulated combat units or human-in-the-loop interactors. This has a big impact in reducing the burden of supporting simulations needed for Dstl studies. Two stochastic optimisation Artificial Intelligence (AI) techniques have been used (Genetic Programming and a novel implementation of Simulated Annealing). The algorithms have been employed in a generic architecture that allows simple application to different problems. This is central to the approach taken. The problem considered is the Dstl level land engagement simulation, SimBrig. This is a highly detailed and complex simulation, with execution times that prohibits the many runs required for the AI to consider a wide range of possible solutions (in this case orders for the military units under AI control). Also, application of stochastic optimisation AI directly on this model would result in solutions that exploited the detailed complexities of the engagement simulation, rather than were based on sound tactical reasoning. However, being able to switch the AI between problems means that a solution can be quickly generated against a simplified wargame (a meta model) which represents only the essential elements of the full wargame problem, whilst still referencing the full simulation (SimBrig) to evaluate the quality of the solution as necessary. The meta model is designed to be quick and robust, without the complexities that would be exploited by the AI. However all essential elements of the tactical problem are modelled (albeit as simply as possible) so that the solutions considered by the AI are properly evaluated. A novel approach has been taken in the way the problem is formulated for the AI. Presenting the problem to the AI using military-like syntax results in the AI algorithms efficiently generating plans for tactical problems which resemble human decision making. This paper presents the approach and techniques used in both the AI algorithms and the meta wargame simulation. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://orlabanalytics.ca/jaor/archive/v9/n1/jaorv9n1p11.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
