Adaptive spatial discretization using reinforcement learning

Abstract A well-known challenge for deformation monitoring is the spatial discretization, i.e. choice of points at which measurements are to be taken. Well-chosen employ prior knowledge yield a significant amount information about certain aspect monitored object. However, such set typically made practically expedient or left measurement instrument itself. We aim derive adaptive discretization strategies that implicitly incorporate domain object via cycle interaction and learning. In those strategies, previous impact locations subsequent ones. formulate as decision theoretical problem review framework reinforcement learning formalizes deriving optimal sequential decisions under uncertainty. Iterative algorithms produce solution schemes this control task. benchmark performance compare its results random, pseudorandom, numerically designed on several geodetically motivated examples. Advantages, disadvantages, practical feasibility approach evaluated reveal boost in efficiency data collection scheme compared classical approaches.