Control of fuzzy discrete event systems and its applications to clinical treatment planning

In this paper, we further develop a modeling and control approach to fuzzy discrete event systems that we initially proposed in [LY01, LY02]. We first investigate an optimal control problem in fuzzy discrete event systems. The problem is abstracted from real applications in biomedical fields. The control objective is to maximize a treatment effectiveness measure while keeping some cost below a given level. This problem is difficult because both the effectiveness function and the cost function are state dependent and hence are not monotonic. Furthermore, the state space of a fuzzy discrete event system is infinite in general. We develop an online approach that can solve this problem. We then apply this approach to HIV/AIDS treatment planning, because it is one of the most difficult treatments in medicine. We also develop a novel computerized treatment decision-making system based on the optimal control approach. The preliminary statistic evaluation of our system shows a strong agreement between the physicians and our system in terms of which treatment regimens to be used for patients of various conditions.