Current projects

  • Development of Dynamically Reconfigurable Agent-Based Discrete Event Simulator for Navy Submarine and Ship Training, and Army maintainer/technician training (@ATHENA)
  • Flight scheduling and resource allocation for 725 squadron
  • Optimisation of Recruitment to achieve capability
  • Learning the landscape

Development of Dynamically Reconfigurable Agent-Based Discrete Event Simulator for Navy Submarine and Ship Training, and Army maintainer/technician Training (@ATHENA)

The Australian Defence continually seeks to improve efficiencies and robustness in the aviation training continuum. For example, services may be consolidated into a single location, schools merged or platforms phased in or out of operation. These changes affect both infrastructure and associated governing policies. Further exacerbated by volatile school failure rates, these changes and uncertainties can severely impact the predictability of the graduates to the operational squadrons.

To model such transient system behaviour, while preserving common simulation architecture requires an easily reconfigurable simulation framework capable of explicitly representing the changing infrastructure, domain policies, constraints and flows. Flexibility is required to re-build models and provide a repository of historical changes and finally cost benefits.

The ALGORA team designed a new approach that combines human intelligence and technology to extract more from human talent, employing state of the art Operations Research techniques that provide guaranteed best solutions - not just any solution that works. That is, stable and robust solutions (avoiding shortage or surplus of personnel numbers), providing significant cost savings. Furthermore, the team has employed high-end cognitive visualisation techniques, thus empowering decision makers to make the right decisions at the right time. The result was a dynamically reconfigurable decision support tool (ATHENA), prototyped for management of Aircrew training. This work extends on the findings in the Aircrew domain problem and expands the existing tool for use in the context of Navy submarine and ship traing, as well as Army technician training. The software features include:

1. manpower tracking – high-end representation of live manpower flows, intuitive diagnostics for identifying bottlenecks, underflows, excessive queuing and more

2. prediction or forward forecasting using hybrid discrete event and agent based simulation, novel use of active and passive probing for monitoring or goal setting

3. strategic what-if scenario analysis, using design space exploration with depth-then-breath strategy

4. optimal adaptive control – ability to self repair the open loop system! Given the current state of the training system, recommend the optimal recruitment policy that guaranties achieving capability, whilst also being a stable solution over long term

5. transparent visualization for all users with functionality catering for broad spectrum of ranks and positions

Flight scheduling and resource allocation for 725 squadron

725 Squadron currently employs a number of staff to manage the squadron scheduling function providing a daily program to achieve squadron objectives. Within the scope of the aircrew training analysis activity, the ALGORA Systems Team is aiming to develop a scheduling and planning tool to optimise and automate much of the process.

The objective of the work is to provide an answers on sufficiency of physical and human resources, in 725-squadron, in order to achieve the target annual graduation numbers. A high fidelity simulation was developed to virtually replicate the operations of the 725 squadron capturing on a minute-by-minute basis, movements of all students (pass/fail/repeat), as they are progressing through their training competencies and competing for the available resources, whilst respecting the squadron business rules. 

The team is currently investigating a number of strategies for producing weekly plans/timetables, given students, syllabi, physical resources and instructors, with business rules acting as constraints. The first version of the Integer Linear Programming algorithm has been developed (CPLEX ILOG and C#), and is currently under testing. Heuristic approaches are also being considered.

P. Lalbakhsh, V. Mak-Hau,  R. Seguin, V. Nguyen, A. Novak, Capacity Analysis for Aircrew Training Schools – Estimating Optimal Manpower Flows Under Time Varying Policy and Resource Constraints, Proceedings of WinterSim 2018, Gothenburg, December 2018

Optimisation of recruitment to achieve capability

The training continuum from recruitment agencies through a complex course structure to production of highly trained pilots, observers or air crewmen is driven by operational needs of the squadrons. In this project, recruitment across courses is optimised to achieve long term squadron capability at minimal cost. Represented as a Markov Decision Process and solved using Dynamic Programming a long term optimisation policy is able to gain significant savings. Currently a simplified prototype system has been produced that will, in the near term, be extended to cover the entire training continuum. Significant challenges in computational complexity will need to be address.

The anticipated outcomes of this work include: optimal policy, cost trade-offs, and “what-if” analysis.

1. S. Pike, B. Moran,  D. Kirszenblat, A. Novak, A Stochastic Programming Approach to Optimal Recruitment in Australian Naval Aviation Training - “The Delta Model”, Proceedings of WinterSim 2018, Gothenburg, December 2018

2. B. Hill, D. Kirszenblat, B. Moran,  A. Novak, Optimising Recruitment to Achieve Operational Capability Conditional on Appetite for Risk, Proceedings of WinterSim 2018, Gothenburg, December 2018

3. S. Suvorova, A. Novak, B. Moran, T. Caelli,  Optimal recruitment strategies for Australian Naval Aviation training using Linear Programming and Dynamic Programming solutions - ''The price of thinking short term", Journal of Military Operations Research (2017), MOR1711 (accepted)

Learning the landscape

Optimisation methods are at the core of much of the work of the ALGORA Systems Team. Fundamental studies in a range of optimisation methodologies are in process to enhance and extend our use-driven research and development. One study is considering new methods for learning and modelling the fitness landscape based on Bayesian ideas. Related issues such as the need to understand how various constraints contribute to the convergence of optimisation algorithms are also being explored.