Online Suspension Parameter Estimation for Improved Road Roughness Estimation

Sammanfattning

The aim of this thesis is to investigate the feasibility of improving road roughnessestimation using only vertical acceleration measurements in the presence ofunknown suspension parameters. An additional objective is to evaluate whethersuch estimation can be performed online, and whether reliable road roughnessestimation can be achieved across vehicles with different suspension systems.The study first evaluates the performance of a previously developed road roughnessestimation method under different suspension configurations in order to assesshow suspension characteristics affect the estimated International RoughnessIndex (IRI). Based on the identified limitations, an adaptive estimation frameworkis proposed to reduce the dependency of the estimated road roughness onvehicle-specific suspension properties.A quarter-car model with augmented suspension parameters was used, and bothan Extended Kalman Filter (EKF) and an Iterated Extended Kalman Smoother(IEKS) were applied for joint state and parameter estimation. The methods wereevaluated using pre-processed vehicle measurements collected under various drivingconditions.The results show that the available measurements generally do not provide sufficientexcitation for reliable estimation of the suspension parameters, resulting inpoor convergence and parameter drift. To mitigate this issue, physical constraintswere imposed on the parameter estimates. Despite the limited observability ofthe suspension parameters, stable and accurate International Roughness Index(IRI) estimation was achieved.Furthermore, the EKF demonstrated computational efficiency suitable for realtimeimplementation, whereas the IEKS provided improved robustness againstparameter drift at the cost of higher computational complexity. The proposedadaptive framework reduced the variation in IRI estimates between differentsuspension systems, although this improvement was achieved at the cost of aslight reduction in estimation accuracy compared with the previously developedmethod. Overall, the results indicate that reliable online road roughness estimationis feasible using vertical acceleration measurements, provided that limitationsrelated to system excitation are appropriately handled.