Homayounzade Mohamadreza, Sajadian Shayan
Mechanical Engineering Department, Fasa University, Fasa, Iran.
IET Syst Biol. 2025 Jan-Dec;19(1):e70019. doi: 10.1049/syb2.70019.
In this paper, we propose a robust control method for the automatic treatment of targeted anti-angiogenic molecular therapy based on multi-input multi-output (MIMO) nonlinear fractional and non-fractional models using the backstepping (BS) approach. This protocol aims to eradicate tumour cells while preserving high levels of the body's natural effector cells and maintaining drug dosage within safe limits. The exponential stability of the controlled system is mathematically demonstrated using the Lyapunov theorem. Consequently, the tumour volume's convergence rate can be precisely controlled-a critical factor in cancer treatment. To fine-tune the controller gains, a soft actor-critic (SAC) algorithm within the framework of deep reinforcement learning (DRL) is employed, with a reward function designed based on the specific requirements of the system. Additionally, the Lyapunov theorem is used to mathematically verify the system's robustness against parametric uncertainty. Compared to state-of-the-art approaches, the proposed scheme demonstrates superior long-term performance, achieving complete tumour eradication and drug delivery convergence to zero within 50 days while preserving high effector cell levels.
在本文中，我们提出了一种基于多输入多输出（MIMO）非线性分数阶和非分数阶模型的稳健控制方法，用于靶向抗血管生成分子疗法的自动治疗，采用反步（BS）方法。该方案旨在根除肿瘤细胞，同时保持体内高水平的天然效应细胞，并将药物剂量维持在安全限度内。利用李雅普诺夫定理从数学上证明了受控系统的指数稳定性。因此，肿瘤体积的收敛速度可以得到精确控制，这是癌症治疗中的一个关键因素。为了微调控制器增益，在深度强化学习（DRL）框架内采用了软演员-评论家（SAC）算法，并根据系统的特定要求设计了奖励函数。此外，利用李雅普诺夫定理从数学上验证了系统对参数不确定性的鲁棒性。与现有方法相比，所提出的方案表现出卓越的长期性能，在50天内实现了肿瘤的完全根除和药物递送收敛至零，同时保持了高效应细胞水平。
