Improving the state estimation for optimal control of stochastic processes subject to multiplicative noise