This paper proposes a modified robust model predictive control (MPC) scheme for linear parametric variables (LPV) or hybrid systems based on a quasi-min–max algorithm. One of the shortcomings of existing algorithms is the occurrence of undesirable disturbances in the control system during dynamic switching (especially in random switching). Also, the control output of other algorithms starts with a non-zero value, which results in practical issues in the experiment. Introducing a new cost function definition, we were able not only to produce control with zero initial value but also to reduce or even eliminate the unwanted disturbances during switching. In addition, the effects of uncertainties are reduced in the prediction dynamics, thereby improving the output signals’ convergence to the desired output as well as the system's robustness. By using the linear matrix inequalities (LMI), we first employed an off-line robust state observer for LPV systems, which rigorously performed even in the presence of random switching. Afterwards and by means of LMIs, it is shown how to construct a modification and improvement in the online optimization problem for the robust output feedback MPC scheme with constraints. Maintaining the system robustness, the proposed modified MPC scheme guarantees the robust stability of the piecewise linear switching parametric variable systems subjected to input constraints. The characteristics and features of the proposed modified MPC scheme are investigated through a numerical example.