InternNav – Navigation Toolbox (github.com)

InternNav has launched an open-source navigation toolbox designed for embodied navigation using PyTorch, Habitat, and Isaac Sim. This comprehensive toolbox allows modular customization of navigation systems, including vision-language navigation and visual navigation towards various goals. Its compatibility with major simulation platforms makes it a versatile tool for researchers and developers. InternNav supports an array of datasets and benchmarks, providing over six comprehensive datasets and ten popular baselines, including the state-of-the-art InternData-N1, featuring more than 3,000 scenes and 830,000 VLN data entries. This significant development stands out due to its introduction of the InternVLA-N1, a dual-system foundation model that achieves superior performance across multiple navigation benchmarks and boasts zero-shot generalization capabilities in real-world scenarios. With features like Interactive Instance Goal Navigation (IIGN), and the VL-LN Bench for evaluating navigation tasks, this toolbox paves the way for advanced research and application in AI-driven navigation systems. By fostering community engagement through an open-source platform, InternNav aims to enhance collaboration and innovation in robotic navigation solutions across diverse environments.