3D Semantic Mapping and Navigation for Robot Task Planning

Ongoing 2025 – present

Semantic 3D maps that let robots find objects and plan where to go.

We build 3D semantic maps that capture objects, furniture and rooms, and split each map into zones such as kitchen or living room. Robots use these maps to navigate to the right place and to pick and place objects. The system is developed in simulation (OmniGibson/BEHAVIOR and ManiSkill) and demonstrated on a real quadruped robot with two robot arms. In collaboration with ETRI.