This paper deals with manpower planning using a dynamic and interactive simulation system that is agile and adaptive to robustly accommodate change — without requiring a complete rewrite. The simulation architecture extends the current hybrid modelling paradigm which integrates agent based (AB) constraints and controls with a discrete event simulation (DES) methodology. This allows for a more expressive, authentic representation of both process flows and agent policies that captures the advantage of system dynamics (SD) modelling by integrating agile controls with response feedback. This approach is inspired by the need to develop an aircrew training pipeline simulation for the Australian Defence Force (ADF) that supports the real needs for strategic manpower planning in a context of policy and requirements change management. A case study is provided to illustrate the challenges and approach.The Australian Defence Force is continually looking for greater efficiencies in its training and operations. These include infrastructure consolidation, new initiatives, reconfigurations as well as policy changes — all having new demands on manpower requirements and flows. For example, basic helicopter flight training components across multiple services are currently being consolidated into a single location. Other types of changes include phasing out platforms, introducing new technologies, consolidating schools, restructuring training facilities and other training resources. Changes like these are common, illustrating that the system is in a constant state of transience and that inevitably causes perturbations in the flow of students through the training program.
However, these infrastructure changes are not the only changes in defense. The policies that govern training in individual schools and squadrons are also susceptible to change. For example, changes in school pass rates can severely impact on the predictability of graduate supply to operational squadrons. Equally, sudden changes in workforce attrition have implications on capability sustainment and recruitment needs.
Whether the changes are driven by internal initiatives or are imposed by the changes in the environment, the decision makers need to respond to these changes and insure both smooth transitions as well as longer term sustainment strategies. Clearly, there is a need for an organization such as the ADF which operates in dynamic and complex environments to create and maintain internal knowledge for better decision making (Baškarada et al. 2016).
This work is motivated by the need for a decision support environment capable of overcoming the challenges posed by dynamic nature of infrastructure and policies in defense. In particular, this paper discusses a user-driven solution to manpower planning, designed to be agile in addressing the needs of the ADF. These include current and projected view of the training system at hand, as well as the ability to perform strategic what-if analyses, to improve the efficiency and operational effectiveness. To model such changes, while preserving a common simulation architecture, there needs to be an easily reconfigurable simulation architecture capable of explicitly representing the changing infrastructure, domain policies, constraints and flows. This approach has the benefit of decreased effort in re-building models while providing a repository of historical changes that all translate into significant cost benefits.
The paper is organized as follows. Section 2 presents related work, while Section 3 describes the simulator design, covering design considerations leading to the proposed architecture. Implementation, verification and validation of the simulator is discussed in Section 4, followed by the detailing of a case study in Section 5. Section 6 concludes the paper.
Agent Based-Discrete Event Simulation Architecture
The conceptual model of the manpower training continuum utilizes both discrete event and agent-based simulation approaches, whilst encapsulating the lever control and feedback loop elements that were previously modelled using the system dynamics approach (Johnstone et al. 2015). The simulator’s architecture is illustrated in Figure 1.