GBB Embodied Interaction AI InstituteGBB / EMBODIED INTERACTIONCollaborate As models begin to perceive, collaborate and act,
we study how they relate to the physical world.
Language models can describe the world. In embodied settings, location, time, action costs and human intent also shape decisions.
GBB focuses onrepresentation, coordination and action loops: preserving decision evidence while allowing people to redefine goals.
From research questions to system structureExtend vision and language into representations of location, constraints and tasks.
Study task decomposition, role boundaries, shared state and conflict resolution.
Make intent clarification, human confirmation and interruptible execution part of interaction design.
Scroll through the cognitive layers. Each answers a distinct question, distributing responsibility across the system.
Bring vision, task language and device events into a common context. Preserve timestamps and sources so expired states are not mistaken for current facts.
Use task-relevant states to distinguish observations from predictions. Every candidate action needs explicit preconditions.
Observation, planning, execution and review carry distinct responsibilities. Share task states with clear provenance and route conflicts through explicit arbitration.
Observe the environment again after execution. Failures, interruptions and human corrections become inputs to the next plan.
Draw obstacles and move endpoints across 1,440 cells. Compare search strategies and watch the route adapt.
Specify the task, inputs, environment and prohibited actions before making broad capability claims.
Compare single-role and multi-role workflows while separating environmental changes, strategy changes and human intervention.
Retain unreachable tasks, conflicting inputs and interruption logs as part of the research record.
Validate interactions through focused prototypes before progressing to model integration and physical device testing.
Why embodied representations need positions, goals and constraints together.
A research framework for shared state, task decomposition and conflict resolution.
From action logs to traceable environmental feedback.
Work with research teams, technology partners and site owners to define tasks, validation environments and interaction prototypes.
Discuss embodied intelligence, task coordination and physical interaction. Submitted details are stored by this website to handle your collaboration inquiry.