VLingNav is a VLA model for embodied navigation that combines an adaptive chain-of-thought mechanism (AdaCoT) to trigger explicit reasoning only when needed with a visual-assisted linguistic memory module (VLingMem) to build persistent cross-modal semantic memory for long-horizon navigation.

We introduce a cross-embodiment and cross-task Navigation Foundation Model (NavFoM), trained on eight million navigation samples that encompass quadrupeds, drones, wheeled robots, and vehicles, and spanning diverse tasks such as vision-and-language navigation, object searching, target tracking, and autonomous driving.

TrackVLA++ is a novel Vision-Language-Action model that incorporates spatial reasoning and target identification memory, enabling superior performance in both long-horizon and highly crowded tracking scenarios.

We present TrackVLA, a vision-language-action model capable of simultaneous object recognition and visual tracking, enabling zero-shot deployment in real-world scenarios.

We present Uni-NaVid, the first video-based vision-language-action (VLA) model designed to unify diverse embodied navigation tasks and enable seamless navigation for mixed long-horizon tasks in unseen real-world environments.

A novel spatiotemporal calibration toolkit for event-based and frame-based camera system.

A spatially compact visual navigation system for autonomous OMS (Oral and Maxillofacial Surgery) suture robot.

A visual marker based on projective invariance called CylinderTag is developed, which provides a new solution for high-precision position estimation of curved objects.

A subpixel corner detector for curved checkerboard-like markers based on quadratic form.