Learning air combat parameters using differential evolution
2013 (English)In: Frontiers in Artificial Intelligence and Applications: Twelfth Scandinavian Conference on Artificial Intelligence, IOS Press, 2013, 225-234 p.Conference paper (Refereed)
In this paper, we improve the performance of autonomous air combat behaviours using a differential evolution approach. We have previously developed air combat behaviours, that are used as opponents in a pilot training facility of the Swedish Air Force. These behaviours contain a number of design parameters, such as at what distance to fire missiles and what risk levels toaccept before disengaging. We improve the performance of the behaviours by applying two approaches. First, by searching for the best response against a given opponent strategy, second by looking for a Nash-equilibrium in a game where both players are allowed to adapt their strategies. Differential Evolution is used in both cases, and the fitness of the parameters is evaluated using parallel execution of detailed high fidelity models of aircraft, weapons and behaviours.
Place, publisher, year, edition, pages
IOS Press, 2013. 225-234 p.
, Frontiers in Artificial Intelligence and Applications, ISSN 0922-6389 ; 257
Differential Evolution, Game Theory, Nash Equilibrium, Training Simulations
Other Computer and Information Science
IdentifiersURN: urn:nbn:se:kth:diva-147495DOI: 10.3233/978-1-61499-330-8-225ISI: 000343477100024ScopusID: 2-s2.0-84894652086ISBN: 978-161499329-2OAI: oai:DiVA.org:kth-147495DiVA: diva2:730291
Twelfth Scandinavian Conference on Artificial Intelligence, November 2012, Aalborg, Denmark
QC 201406272014-06-272014-06-272014-11-21Bibliographically approved