As mobile robots are deployed in increasingly com- plex domains in the open world, the level of detail demanded by the robot’s decision-making model to ensure reliable operation increases. To support this, mobile robotic agents are fixed with a wider array of more informative sensor equipment and downstream perception systems that convert such sensors’ information into usable representations by the agent’s planning models. Automated decision making models often assume free and consistent access to such information. However, in the con- text of a mobile robot, this assumption may not hold, and failing to account for this in the plan- ning model may lead to costly behavior or even failure. In this paper, we propose a mixed open- loop/closed-loop planning model based on memory states that integrates knowledge about limitations on sensory feedback in order to proactively plan around these limitations or exploit situations where costly sensing is unnecessary. We provide both the- oretical properties as well as empirical evaluations on a simulated mobile robot domain.