Title: Blood pressure regulation with RAS model and Q-learning control
Authors: Ciprian Sandu
Addresses: Faculty of Automatic Control and Computers University Politehnica of Bucharest, Splaiul Independenţei 313, Bucureşti 060042, Romania
Abstract: This paper focuses on the blood pressure regulation in post cardiac surgery patient, specifically on showing that Q-learning (a reinforcement learning method) is suitable for the automatic control of blood pressure in post cardiac surgery patients. It uses directly the clinical data; there is no need of a patient model. Such a model is used however, but only as a source of artificial clinical data. Firstly, we use this model in order to create closed-loop control with a classical PID controller and then with polynomial RST control (where a robustness analysis is performed). Secondly, as the contribution of this paper, we use the model only as a source of pseudo-clinical data with reinforcement learning, the Q-learning algorithm. In all these cases, the model contains the RAS reflex model of the body.
Keywords: blood pressure regulation; reinforcement learning; Q-learning; mathematical modelling; renin-angiotensin system; RAS reflex model; PID control; polynomial RST control; post cardiac surgery; heart surgery; clinical data.
International Journal of Advanced Intelligence Paradigms, 2017 Vol.9 No.1, pp.67 - 81
Received: 12 Jun 2015
Accepted: 03 Jul 2015
Published online: 26 Dec 2016 *