Long-Horizon Navigation is the ability of embodied agents to efficiently explore, map, and navigate complex environments over extended temporal and spatial scales. This involves constructing a topological representation of the scene, leveraging motion planning to reach specified waypoints, and employing object navigation strategies to locate and transport semantic targets. The modular hierarchical transport policy enables generalization to more challenging environments by training on simplified tasks, ensuring adaptability. Additionally, long-horizon navigation incorporates deep exploration and planning to optimize object retrieval and transport while managing load constraints and variable capacities.