Humanoid Robots Can Now Carry Boxes Through Cluttered Rooms
This is the skill robots need before they can tidy your home or stock shelves.
Human-shaped robots are getting good at walking. What they're still bad at is the thing delivery workers do all day: carrying something through a crowded room without knocking things over. A new research paper introduces HOTICE, a training method that lets a humanoid robot walk and carry at the same time while steering around clutter. The team tested it both in computer simulation and on a real Unitree G1 humanoid — a roughly human-sized robot you can buy today.
Two ideas make it work. First, the system imagines invisible "push-away zones" around both the robot and the object it's holding, so the robot guides both of them around obstacles instead of only worrying about itself. Second, it splits the robot's control into a lower body (legs, balance, walking) and an upper body (arms, torso), then keeps the two halves talking so the whole body moves as one. The team also had specialist trainers teach a single general-purpose robot, a common trick for making AI handle unfamiliar situations.
Why does this matter to you? Carrying stuff is the unglamorous skill that decides whether a robot is genuinely useful or just a demo. If a robot can haul a laundry basket past a coffee table, or move boxes down a cramped warehouse aisle, it starts to replace real labor — warehouse shifts, elder-care errands, hotel room service. It also matters for safety: a machine that ignores the objects around it is a hazard in any space where people work.
The catch: this is research, not a product. It was proven with one robot model in controlled tests. Real homes have pets, kids, stairs, and unpredictable messes. There's no price tag, no release date, and no guarantee it works outside the lab yet.
- HOTICE lets a humanoid robot carry an object while also avoiding obstacles around it — the robot plans for both itself and the thing it's holding
- It was tested on a real Unitree G1 humanoid, not just in simulation, which is a meaningful step toward practical use
- The robot keeps its legs and upper body coordinated by training them separately but sharing information, so it balances while carrying awkward shapes
Why It Matters
Carrying objects through messy spaces is the skill robots need before they can help in homes, warehouses, and hospitals.