GalaxyBot has teamed up with research groups from several universities to unveil NavFoM, the inaugural cross-embodiment panoramic navigation foundation model on a global scale. Designed to consolidate navigation tasks across diverse robotic platforms into a cohesive framework, NavFoM facilitates seamless deployment across multiple scenarios. It exhibits exceptional versatility, supporting zero-shot navigation in both uncharted indoor and outdoor settings, eliminating the necessity for supplementary mapping efforts or data acquisition. Additionally, NavFoM accommodates multi-task functionalities, empowering robots of varying dimensions to operate with heightened efficiency within its framework.

From a technical standpoint, NavFoM introduces innovative TVI Tokens and BATS strategies, compiles an extensive cross-task dataset, and doubles the training data volume relative to its predecessors. This landmark release signifies a substantial leap forward in the realm of robotic navigation, providing developers with a robust foundation to craft specialized application models tailored to their specific needs.