Multi-Objective Control To Active Suspensions Based On LMI Optimization

Performance requirements for advanced vehicle suspensions include isolatingpassengers from vibration and shock arising from road roughness (ride comfort),suppressing the hop of the wheels so as to maintain ?rm, uninterrupted contactof wheels to road (good handling or good road holding) and keeping suspensionstrokes within an allowable maximum. Moreover, control inputs, for example activeforces that are generated by hydraulic actuators, are constrained due to actuatorsaturation. However, these requirements are con?icting, for example, increasingride comfort results in larger suspension stroke and smaller damping in the wheel-hop mode. In order to manage the trade-o? between con?icting requirements, many activesuspension control approaches are proposed based on various control techniques,such as LQG, adaptive control, H∞ control and nonlinear control. A commonpoint of most approaches is that all requirements are weighted and formulated ina single objective functional (H2 or H∞ ), which will be minimized to ?nd anoptimal controller. One manages the trade-o? between con?icting requirements bychoosing appropriate weighting, possibly frequency-dependent, which is howevernot trivial. For the 4 DOF half-car model, in order to capture control requirements(ride comfort, good handling, limited suspension stroke and actuator saturation),one may choose a single objective by tuning 8 weights. Clearly, to do this to achievean optimal performance is quite di?cult, if not impossible in practice. Having a close insight into the control requirements for active suspensions, itis clear that requirements on good road holding, limiting suspension strokes andcontrol inputs within bounds are in fact hard constraints in time domain. Hence, itis much nature to formulate the active suspension control problem as a disturbanceattenuation problem with hard constraints. In this thesis we bring forward two IV吉林大学博士学位论文approaches to deal hard constraints: invariable domain constraints control andgeneralized H2 constraints control. In the framework of LMI optimization andmulti-objective control, several mixed control arithmetics, which combining eitherof the constraints control methods and H∞ performance or H2 performance areformulated, respectively. In each of the mixed control arithmetics, design of thecontroller is convert to a LMIbased semi-de?nite programming problem. While onlyone parameter has to be determined and can be used as a free design parameter tomanage e?ectively con?icting requirements. It indicates a big betterment in activesuspension control. The formulation processes and the proofs of above mentioned approaches arepresented in detail in this thesis. Moreover, in order to validate the e?ciencies ofthe proposed approaches, we designed state-output feedback controller to a realhalf car model based on each mixed control strategy. The half car model used isselected unique. To do this, we want to ?nd a set of more suitable mixed criterionscombination in a more fair condition. Design results are analyzed and simulated infrequency responses, time property, pulse responses and μ-analysis. The followingare some conclusions we gained: Constrained H∞ control approach constrained H∞ active suspension systemson half-car models. The H∞ performance is used to measure ride comfort suchthat more general road disturbances than white noise can be considered. It isminimized to enhance ride comfort. Time-domain hard constraints are captureddirectly using the concept of reachable sets and in a state-space ellipsoid de?ned bya quadratic storage function. It is required to keep the time-domain variables withinbounds, but not to achieve a minimum. Hence, the designed active suspensionshave the property to achieve the best possible ride comfort by making the most ofallowable control inputs and suspension strokes, in the precondition of guaranteeinggood road holding. Thus is so c...