An Improved Artificial Electric Field Algorithm for LQR Controller Weight Optimization in Magnetorheological Semi-Active Suspension

Abstract

Magnetorheological (MR) semi-active suspensions can rapidly tune damping for ride-comfort improvement, yet LQR-clipped control remains constrained by a structural mismatch: an unconstrained LQR law generates an ideal force that is subsequently projected onto the feasible MR damper envelope. This paper therefore reformulates LQR weight tuning as an actuator-realizability-aware optimization problem in which body acceleration, suspension deflection, realized control effort and the RMS tracking error between the ideal LQR command and the clipped force are jointly minimized. To mitigate the resulting non-smooth saturation landscape, an artificial protozoa optimizer (APO)-guided improved artificial electric field algorithm (MIAEFA) is developed to leverage local exploitation near actuator-feasibility boundaries. The MR damper is represented at the actuator-boundary level so that the optimization explicitly penalizes force commands that cannot be delivered after clipping. Comprehensive engineering validation is conducted under multiseverity random roads and transient impact excitations. Under ISO Class C random-road excitation at 72 km/h, MIAEFA-LQR reduces body-acceleration RMS by 55.71% relative to the passive suspension and by 15.74% relative to AEFA-LQR, while reducing the force-tracking error between the ideal and actual forces from 172.00 N to 124.55 N. The results show that MIAEFA-LQR circumvents excessive ideal-force demand by aligning LQR commands with the deliverable MR damper range, thereby improving ride comfort through actuator-compatible force realization rather than through larger unrealizable commands.