Multimodal Command Grounding for Telepresence Robot Navigation using Vision–Language Perception and Classical Planning

This framework enables telepresence robots to execute multimodal commands (speech/text and pointing) by grounding them into Semantic Waypoints. It integrates di-rectly with ROS 2 Nav2 to ensure safer, more predictable indoor navigation in dynamic environments such as healthcare facilities, offices, and smart homes.