It is discussed an approach for optimal control of semi‐active vibro‐insulation system. As an object it is accepted a dynamical model of quarter car suspension with magneto‐rheological controlled damper. Based on the quality criteria of insulation there is formed a quadratic functional, which are minimized with the synthesis of a linear quadratic regulator. The essential physical non‐linearity of the controlled damper, lead to serious complications in deducing of needed inverse damper model. For the solution of this, is used appropriate neural network. The numerical results are shown for the concrete model.