A Sliding Mode Controller Based on the Interacting Multiple Model Strategy

Abstract

Sliding mode controllers (SMCs) are a form of variable structure control, which utilizes a discontinuous switching plane along some desired trajectory. This plane is often referred to as a sliding surface, in which the objective is to keep the state values along this surface by minimizing the state errors (between the desired trajectory and the estimated or actual values). Ideally, if the state value is off or away from the surface, a switching gain would be used to push the state towards the sliding surface. This switching brings an inherent amount of stability; and as such, SMCs have become a popular control strategy for systems with modelling uncertainty. Furthermore, some systems behave according to a number of different models (or operating regimes). In these scenarios, it is desirable to implement adaptive estimation algorithms, which ‘adapt’ themselves to certain types of uncertainties or models in an effort to minimize the state estimation error and improve tracking performance. One type of adaptive estimation technique includes the multiple model (MM) algorithm and its interactive form (IMM). For the MM methods, a Bayesian or probability based framework is used. Essentially, based on some prior probabilities of each model being correct (i.e., the system is behaving according a finite number of modes) the corresponding updated probabilities are calculated. This paper introduces a new type of sliding mode controller based on the interacting multiple model strategy (SMC-IMM). The SMC-IMM method is applied on an electrohydrostatic actuator, and the results are compared with the standard SMC strategy.