One advanced means of protecting structures against earthquake ground motions is the use of semi-active devices to customise and limit structural response. Thus, the design, modelling and analysis of different semi-active control devices have received increasing research attention. This study presents a method to determine optimal control forces for magneto-rheological (MR) dampers, using three algorithms: a discrete wavelet transform (DWT), a linear quadratic regulator (LQR), and a clipped-optimal control algorithm. DWT is used to obtain the local energy distribution of the motivation over the frequency bands to modify conventional LQR. The clipped-optimal control algorithm is used to get the MR damper control force to approach the desired optimal force obtained from modified LQR. A Bouc-Wen phenomenological model is utilized to capture the observed nonlinear behaviour of MR dampers. Time history analysis for a single degree of freedom (SDOF) with periods of T= 0.2-5.0 sec is utilized to compare the impact of using classic and modified LQR in controlling the MR damper force under 20 design level earthquakes of the SAC (SEAOC-ATC-CUREE) project. Performance is assessed by comparing the maximum displacement (Sd), total base shear (Fb) and the controller energy. This study shows the proposed modified LQR is more effective at reducing displacement response than conventional LQR. The modified LQR method reduces the median value of uncontrolled Sd by approximately 40% to 88%, over all periods to 5.0 seconds. Moreover, the modified LQR uses about 45% less energy than conventional LQR. Overall, these results indicate the robustness of the proposed method to mitigate structural response and damage using MR devices.