Software Engineering Multi-Agent Simulations for Computational ...

5 downloads 129 Views 253KB Size Report
Oct 1, 2015 - over twenty years experience with the software engineering of multi-agent ... is also discussed. An outline of the issues faced through.
Software Engineering Multi-Agent Simulations for Computational Operations Research: Challenges and Lessons from 20 Years of Practice Michael Papasimeon Defence Science and Technology Organisation Melbourne, Australia

Kevin McDonald

Defence Science and Technology Organisation Melbourne, Australia

Extended Abstract This paper presents challenges faced and lessons learnt from over twenty years experience with the software engineering of multi-agent simulations for computational operations research of air to air combat in the Defence Science and Technology Organisation (DSTO). A major role of the DSTO is to undertake research and development to support Australia’s current and future air combat capability. Operational analysis is a key part of this research and development program, and is used to support multi-billion dollar fighter acquisitions, systems upgrades and the assessment and evaluation of air combat tactics. Computational modelling and simulation, and in particular multi-agent simulations of air combat, are integral tools in many operational analysis studies of air operations. Multi-agent simulations of air combat include computational representations of aircraft flight dynamics, sensors, weapons, communications, countermeasures and electronic warfare systems. They also include representations of pilot and team decision using intelligent agent reasoning architectures. These simulations are executed many thousands of times building up a statistical picture of key operational analysis measures using Monte-Carlo methods. This is often known as a constructive simulation where the computer is pitting one group of fighter aircraft against another in a virtual battlespace In e↵ect operational analysts undertake computational science using these systems as computational models of a complex socio-technical system; in this case the air combat domain, exploring possibilities and what-if scenarios in order to gain insight into specific air combat problems.

Lyndon Benke

Defence Science and Technology Organisation Melbourne, Australia

challenges but have also yielded significant lessons. Lessons regarding requirements capture and system design, implementation, test, verification and validation are described. More importantly the key engineering challenges faced, how they were overcome and how the lessons learnt have been incorporated into developing the next generation of multiagent simulation are presented. The software engineering story in this paper starts in the late 1980s and early 1990s and presents five generations of multi-agent simulations starting with procedural systems that evolve along the way to hybrid object and agent oriented systems. It is a story about software engineering that sits at the intersection of scientific computing, intelligent systems, air combat, operations research, computational modelling and simulation. The challenges of introducing software engineering tools, techniques and methods to a group of scientists and engineers undertaking ad-hoc programming activities without software engineering rigour is discussed. The advantages and disadvantages of running a parallel research program into artificial intelligence, intelligent agents and cognitive science the outputs of which are then incorporated into successfully deployed applications to support client programs is also discussed. An outline of the issues faced through the integration of legacy code often sourced from disparate provenances is presented. The paper also canvasses the key methodological and practical issues of understanding a domain as multi-faceted and complex as air combat.

The software engineering of these class of multi-agent simulations over the past twenty years have presented significant

Learning to overcome these challenges has led to many software engineering lessons. These include lessons regarding the size of teams, understanding requirement complexity, traceability, collaborating with other organisations, the suitability of di↵erent programming languages and tools for different tasks and undertaking software development in an analysis focused organisation.

Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for thirdparty components of this work must be honored. For all other uses, contact the Owner/Author. Copyright is held by the owner/author(s). ASWEC ’ 15 Vol. II, September 28 - October 01, 2015, Adelaide, SA, Australia ACM 978-1-4503-3796-0/15/09. http://dx.doi.org/10.1145/2811681.2817753

We conclude by describing the key software engineering lessons from the past twenty years of development that have been adopted to influence the philosophy and engineering processes for current and future systems. Specifically we describe the Air Combat Environment (ACE), a fifth generation air-to-air combat multi-agent simulation currently in development and how the lessons and challenges of previous generations are driving the technology for future air combat constructive simulation.

123

Suggest Documents