Adaptive PID control using setpoint classification: a lightweight and interpretable alternative to intelligent controllers

This paper proposes a novel setpoint classification-based adaptive PID controller for improving transient and steady-state performance in time-delay and unstable process systems. Building on the dynamic setpoint weighting framework, the proposed method introduces a rule-based mechanism that adjusts the proportional and derivative setpoint weighting parameters (β,γ) in real time, based on the classification of the reference signal type: step, ramp, or drift. This classification enables the controller to automatically adapt its behavior to match the nature of the desired trajectory, thereby enhancing responsiveness and robustness. The method is validated using comparative simulations against both classical and modern intelligent control approaches, including IMC-based PID, neuro-Fuzzy PID, and a reinforcement learning-based PID using TD3. The results demonstrate that the proposed method achieves superior performance in terms of integral error metrics (IAE, ISE), while maintaining simplicity, interpretability, and practical deployability, making it well-suited for industrial applications.