An Explorative Study of Search Algorithms Applied to the Gain Scheduling in PID Controllers

The efficiency of a proportional-integral-derivative (PID) controller depends directly on the values of the gains
$K_{p},K_{i}$
, and
$K_{\mathrm{d}}$
. In this context, changes in the process dynamics imply degradation in the performance of the closed-loop control system. Several artificial intelligence techniques are used to adjust the PID parameters to improve the control system's performance. However, there is little exploratory evidence about how primitive techniques with low computational costs can contribute to solve this gain calculation task. This paper explores and compares three algorithms applied to the real-time gain calculation of PID controllers: fuzzy logic (FGS-PID), simulated annealing (SA-PID), and the A * search method (A *-PID). The validation of the operation of these schemes is carried out through simulations as well as by applying these controllers to a two-tank liquid-level control system. Experimental results show the algorithms' effectiveness to improve the control system's global performance and to add passive fault-tolerance capabilities.