Simplified decision making in the belief space using belief sparsification

In this work, we introduce a new and efficient solution approach for the problem of decision making under uncertainty, which can be formulated as in belief space, over possibly high-dimensional state space. Typically, to solve problem, one should identify optimal action from set candidates, according some objective. We claim that often generate an analogous yet simplified solved more efficiently. A wise simplification method lead same selection, or maximal loss optimality guaranteed. Furthermore, such is separated inference does not compromise its accuracy, selected would finally applied on original state. First, present concept general problems provide theoretical framework coherent formulation approach. then practically apply these ideas by considering sparse approximation their initial belief. The scalable sparsification algorithm able yield solutions are guaranteed consistent with problem. demonstrate benefits realistic active-SLAM manage significantly reduce computation time, no quality solution. This work both fundamental practical holds numerous possible extensions.