The problem of estimating the kinematic state (position and velocity) of a moving emitter by the use of angle-of-arrival (AOA) information obtained at times t0, t1 and so on, is discussed. For simplicity, the discussion is restricted to planar (i.e., two-dimensional) motion. The set-valued Kalman filter provides a family, or set, of all track estimates that are consistent with the observed signals and all available contextual and logical information regarding initial conditions of the emitter. The Kalman filter propagates one state estimate: the one derived from a specific choice of a priori state. A natural extension of this estimator is to specify that the initial conditions lie in a convex, bounded region of state space, and to propagate this entire region, rather than just one point. The resulting estimator is a generalization of the well-known Kalman filter.