The stability analysis of reset control systems is challenging when reset time instants are uncertain, because such uncertainties may damage or even destabilise the system. This study considers reset control systems with uncertain output matrix, and presents sufficient conditions for the quadratic stability and finite gain L 2 stability of the closed-loop reset systems. Then, the results are extended to piecewise quadratic stability, which is much less conservative. The design of the reset controller is also discussed. All the results are given as linear matrix inequalities by using a lemma about multi-convex function. Several examples are given to show the effectiveness of the obtained results.
IntroductionReset controllers were first proposed in [1], where the so-called Clegg integrator was introduced, the controller state is reset to zero when its input and output satisfy some pre-defined conditions. In [2], the Clegg integrator was generalised to first-order reset element (FORE), where it was suggested that introducing reset control may overcome some fundamental limitations and improve transient performance of linear control systems. An explicit example was given in [3] to say that some specifications can be met by reset controller but not by any linear time invariant (LTI) controllers, more explicit examples can be found in [4]. The FORE with zero crossing reset conditions was used to improve transient performance such as in [5][6][7]. Delay-dependent and delay-independent stability results of reset systems were given in [8-10], respectively. Well-posedness of reset control systems with and without reset band were considered in [11,12]. In [13,14], dissipativity theory was used for stability analysis of reset control systems. More results about zero-crossing resets can be found in [15] and references therein. In [16], a new model of FORE under the hybrid systems framework of [17] was proposed that allows jumps on more complicated sets, and temporal regularisation is introduced to avoid Zeno solutions. With this new model, the exponential stability and input-output stability were presented, and these results provide a systematic design tool for reset controllers. The L 2 stability result for reset control systems was first presented in [18]. Then, linear matrix inequality(LMI) conditions were given in [16] to estimate the L 2 gain and piecewise quadratic Lyapunov functions were used to get a tighter L 2 gain estimation. In [19] the analytical and numerical Lyapunov functions were provided to prove stability and L 2 gain performance of FORE control systems. Furthermore, the LMI-based analysis method was used to determine the performance of single-input-single-output (SISO) reset systems in both L 2 gain and H 2 sense as discussed in [20]. Recently, improved reset rules were proposed in [21] such that strictly decreasing Lyapunov functions at jumps can be constructed.Most of the above mentioned literature considered the control structure in Fig. 1 with certain system matrices. The control input signal is u and ...