Robotic Cellular Warehousing Systems provide an innovative robot-to-goods picking approach designed to improve robot transportation efficiency, where robots move to pick items and transport the picked items to workstations. In this study, we investigate the optimal operating policies for such a system by comparing two picking strategies (pick-while-sort and pick-then-sort) and three robot-to-workstation assignment rules (random, closest, and dedicated). Specifically, we develop dedicated closed queuing networks to model robot-to-goods picking and estimate warehouse throughput under different policies through single-class and multi-class models. The effectiveness of these analytical models is validated through numerical simulations, with an average gap of 5.53% between simulation and analytical results. Additionally, we conduct a series of numerical experiments to examine the impact of various factors on warehouse performance, including the numbers of robots and workstations, robot capacity, order size, and sorting efficiency. Based on the experimental findings, we provide managerial implications that offer insights into optimizing resource allocation and system configuration. These insights enable warehouse managers to improve operational efficiency and overall performance.