A reinforcement learning development of the FRAM for functional reward-based assessments of complex systems performance

Abstract

Although the Functional Resonance Analysis Method (FRAM) is a well-established approach to visualizing complex systems' operations in terms of functions, further improvements are required to examine systems' performance through functionality. This study aims to develop an approach to couple the FRAM to reinforcement learning (RL) to explore complex operations. The developed approach is called the functional RL approach and constitutes a novel way of using a FRAM model to explore functionality using an artificial intelligent (AI) agent who responds to reward values assigned to functional parameters. To exemplify the approach, an agent is employed to perform the role of a patient and explore a functional environment generated by the FRAM. Reward values are considered to motivate the agent in order to explore the environment to achieve its objective. The ability of the developed approach is examined using different scenarios implemented in healthcare operations. The results of using the functional RL approach indicate that the approach is able to specify the functional route taken by the agent and to examine the performance of the system based on accumulated rewards. The outcomes of this study demonstrate that the developed functional RL approach provides a novel means to explore operational environments to identify the routes that have potential to affect the system performance. This method can be used as a powerful way to assess how a system performs under different management structures.