This Project Grant from the National Science Foundation's Computer and Information Science and Engineering program ($199,481) supports research at George Mason University from June 2023 through May 2026 to develop an approach for improving service robot decision-making and behavior during deployment in unfamiliar environments. The award aims to facilitate fast and reliable online retraining and adaptation of robot planning despite missing world knowledge. The researchers will develop a principled method for improving robot performance over long-term planning tasks in partially mapped settings, emphasizing reliability, data efficiency and outcomes. This will be achieved by coupling data-driven and classical planning approaches, allowing learning to augment model-based planning and enable completeness despite knowledge gaps. The robot will leverage online experience collection and expert guidance to self-audit, retrain learned behaviors, and improve long-term deployment performance for complex multi-stage service robotics tasks.