Knowledge of the system state is key in autonomous decision making frameworks. To meet the run-time requirements and memory limitations that apply in this context, we propose an incremental estimation strategy that limits the number of estimates at each time step while still guaranteeing that dead-ends are not encountered. As this is not always achievable with a single-state estimator, we increase the number of estimators by defining the new notion of multi-estimation. In this paper, we consider discrete event systems and analyse the case of 2-estimation, i.e. taking two single-state estimators to form a 2-estimator. We first present a necessary condition for a 2estimator being dead-end free and then derive a necessary and sufficient condition, both illustrated by experiments.