Self-organising learning control and its application to muscle relaxant anaesthesia

The concept of a self-organising control system is attractive in biomedicine because of the imprecise nature of available physiological models. In this paper a particular strategy called a self-organising controller (SOC) originating from the work of Barron on aerospace systems is applied to the control of muscle relaxant anaesthesia. The SOC algorithm, which requires no prior knowledge of system dynamics, is described, both in single variable and multivariable format. Simulation results are presented for SOC performance on a well-established pancuronium model. Three implementations are described, being the use of a general purpose language, a SUN workstation approach, and a parallel computer transputer solution. The latter approach becomes important for multivariable control because of the computing-intensive nature of SOC. The transputer is shown to be a suitable vehicle for implementation in terms of speed and parallelism for SOC.