[en] H_∞ optimal estimation theory is formulated as a minimax problem where the maximum 'energy' of estimation error over all disturbance trajectories in L₂ functional space is minimized in the sense of H_∞-norm. A design method based on the characteristics of the formulation is presented in which design specifications can be taken into account as frequency weights for the objective function. We applied this method to design a reactivity estimator for a fast breeder reactor. In the design was used a system model which consists of usual reactor point kinetics equations combined with a state equation of net reactivity driven by fictitious noise/disturbance. Finally, we show by computer simulation that H_∞ estimator gives good results in estimating net reactivity changes caused by various disturbances. (author)