Skip to Content
MathematicsMathematics
  • Article
  • Open Access

20 November 2026

49 Pages

Metaheuristic Optimization of PID, FOPID and TIDFDD Controllers for a Half-Car Active Suspension Under Stability and Road-Holding Constraints

Department of Mechanical Engineering, Giresun University, 28200 Giresun, Türkiye
Mathematics2026, 14(22), 4284;https://doi.org/10.3390/math14224284 
(registering DOI)
This article belongs to the Special Issue Developments of Vehicle Dynamics and Intelligent Control

Abstract

Closed-loop stability and road holding are rarely included in metaheuristic tuning of active suspension controllers. A constrained framework is proposed. Each candidate is screened for sampled-data stability and a 6 dB simultaneous gain margin before simulation. Working-space and road-holding limits are penalized. The framework is applied to PID, fractional-order PID (FOPID) and tilt-integral-derivative with filtered derivative and filtered double-derivative term (TIDFDD) controllers. A linear half-car model with actuator dynamics and a 1 ms digital loop is used. The objective combines road tracking and ISO 2631-1 weighted seat acceleration. Six recent metaheuristics are compared with equal budgets per structure, 30 runs per algorithm–controller pair and non-parametric tests. A full-state linear quadratic regulator (LQR), tuned in the same way, serves as a model-based reference. The C/MEX implementation is verified against an independent Simulink model. TIDFDD reduced weighted seat acceleration by 50% on ISO 8608 roads, versus 45% for PID and FOPID. The road-holding constraint was active in all final solutions. With ideal full-state measurements, the LQR achieved a 59% reduction. In the tested delay and parameter-variation scenarios, the PID-family controllers showed smaller performance losses and larger stability margins. The TIDFDD advantage depends on convergence to the better of two observed attraction basins.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.