Design and experimental evaluation of a data-driven PID controller using cerebellar memory

In industrial process control, the proportional-integral-derivative (PID) control scheme is well-recognized and widely utilized. However, due to the distinctive characteristics of real systems, their control design primarily aims at achieving optimal production performance, constrained by uncertainty and variations. This paper initially discusses a database-driven PID (DD-PID) control scheme that was previously proposed. This scheme combines the DD-PID with the cerebellar model articulation control to minimise computational and memory requirements for industrial application. Subsequently, a hydraulic system is introduced, detailing its characteristics and control necessities. Furthermore, both the DD-PID and the proposed cerebellar model articulation control memory-based DD-PID control schemes are implemented and evaluated through experimental examples on a hydraulic system. Lastly, as a practical validation of the theoretical approach, a quantitative assessment compares the two methods, discussing the practicality and efficacy of the proposed scheme in reducing computation and memory consumption. This paper proposes a data-driven controller using a cerebellar memory scheme. It comprises three parts: (i) offline optimisation via database-driven proportional-integral-derivative (DD-PID) control; (ii) data mapping based on similarities of the DD-PID-constructed database; and (iii) real-time implementation via the mapped cerebellar memory. A cerebellar memory based on DD-PID and cerebellar model articulation control can be constructed to reduce the computational and memory consumption of DD-PID control. It also can be implemented to industry-oriented applications even with fast-response time and noise disturbance. image