Research on linear quadratic regulator control strategy for active suspensions based on improved multi-objective coati optimisation algorithm

Weight matrix selection in linear quadratic regulator-controlled active suspensions has long depended on empirical iteration, and single-objective cost formulations cannot simultaneously satisfy the competing demands of vertical ride quality, tyre–road contact and suspension travel. This paper proposes an improved multi-objective coati optimisation algorithm for automated linear quadratic regulator weight tuning. Two enhancements are introduced into the original coati optimisation algorithm: Latin hypercube sampling for initialisation, which ensures even coverage of the search space, and an elite-retention archive based on crowding distance, which keeps the Pareto front evenly distributed. The single-objective variant (improved coati optimisation algorithm) is benchmarked on the IEEE CEC 2022 suite, while the improved multi-objective coati optimisation algorithm is validated on the disconnected-front DTLZ7 problem against the multi-objective coati optimisation algorithm, multi-objective particle swarm optimisation, multi-objective grey wolf optimiser and multi-objective whale optimisation algorithm, where the original multi-objective coati optimisation algorithm fails entirely (HV = 0). Applied to a quarter-car model under Class-D random excitation, the improved multi-objective coati optimisation algorithm-linear quadratic regulator reduces RMS body vertical acceleration by 47.98% and 20.75% relative to the passive and conventional linear quadratic regulator baselines, and suspension travel by 58.75% and 19.89%, respectively, while keeping dynamic tyre deflection within the road-holding safety bound. Robustness is confirmed across no-/half-/full-load and Class-B/D conditions. Since the optimisation is offline and the final controller is a fixed-gain linear quadratic regulator, the scheme suits real-time automotive implementation.