In this paper, we study safe data collection for the purpose of policy evaluation in tabular
Markov decision processes (MDPs). In policy evaluation, we are given a\textit {target} policy …