Can Home Robots Handle Contact-Rich Chores?

That class of work is called contact-rich manipulation. It is where home robots stop being camera-on-wheels demos and start needing force sensing, dexterous hands, whole-body balance, recovery behavior, and training data that captures what contact actually feels like.

The useful buyer question is not "can the robot move?" It is: can it apply the right amount of force, notice when contact changes, and recover before something spills, tears, tips over, or gets dropped?

What are contact-rich chores?

A contact-rich chore is any task where success depends on continuous physical feedback, not just recognizing objects in an image. If the robot only needs to see a cup and drive near it, vision may be enough. If it must grip the cup, lift it, avoid squeezing too hard, keep it level, pour liquid, and set it down on a crowded table, force and touch start to matter.

Households are full of these chores:

This is why a polished thirty-second clip is not enough. Contact-rich chores need many trials across many homes, objects, lighting conditions, furniture layouts, and failure modes.

Why are towels, tea, and cat litter a useful test?

A 2026 research project called Humanoid Touch Dream is useful because its benchmark tasks sound closer to messy household work than the usual humanoid highlight reel. The project tested five contact-rich tasks: Insert-T, Book Organization, Towel Folding, Cat Litter Scooping, and Tea Serving. The team reports that its Humanoid Transformer with Touch Dreaming improved average success rate by 90.9% relative to a stronger baseline, and that predicting latent tactile signals produced a 30% relative gain in part of the ablation study.

Those numbers should not be read as a consumer product claim. They are a research signal. The important takeaway is what the researchers chose to measure: contact changes, tactile prediction, dexterous hands, whole-body control, and real-world rollouts. In other words, the benchmark is asking whether the robot can handle the kinds of physical uncertainty that homes create every day.

That maps directly to newer home-humanoid claims. Figure 03 is listed in the ui44 database as an active humanoid with no public price, a 173 cm height, 61 kg weight, about 5 hours of battery life, and a reported 20 kg payload. Figure's own announcement says Figure 03 adds compliant fingertips, first-generation tactile sensors, palm cameras, improved vision, soft goods, and wireless charging. Those are exactly the hardware clues buyers should look for. They do not prove a household robot is ready, but they show the company understands that home work is about contact, not just motion.

Which robots in the ui44 database are closest today?

No robot in the ui44 database should be treated as a fully general contact-rich home worker today. The better way to compare them is by asking which one has the right ingredients, which one is actually available, and which one is still mainly a research or preorder promise.

This is the trade-off buyers should notice. Stretch 4 is a real product, but at nearly $30,000 it is not a mainstream appliance. 1X NEO is home-focused and soft, but preorder status means buyers should demand evidence before assuming it can do all advertised chores autonomously. Figure 03 has strong hardware language around sensing and tactile hands, but is not priced for consumers. Unitree G1 is available and comparatively low-cost, yet its own product page cautions that the humanoid industry remains early and that individual users should understand the limitations before purchase.

Why vision alone is not enough

Vision-language-action models are improving quickly. They can recognize scenes, parse instructions, and generate motion. But many contact-rich problems are not visible. A cup can look secure while starting to slip. A drawer can look closed while force is building against a jammed runner. A towel can hide the edge the robot needs to grab. A bottle cap can look the same whether a hand is resting on it or squeezing it.

The Robotics & AI Institute makes this point clearly in its discussion of handheld robotic data collection: internet-scale video lacks force and tactile data, egocentric video cannot capture many contact cues, and teleoperation without tactile feedback can lose the subtle force information that makes manipulation work. Its argument is not that cameras are useless. It is that contact-rich manipulation needs data streams that include force, proprioception, and tactile signals, not just pixels.

That is also why DAIMON Robotics' tactile-data push is worth watching. In an IEEE Spectrum interview, DAIMON describes a fingertip-sized tactile module with more than 110,000 effective sensing units, a dataset strategy that includes 10,000 hours of open-sourced data, and a Vision-Tactile-Language-Action approach that treats touch as a core modality. The interview is company-led and should be read with that context, but the technical point is sound: if a robot cannot detect slip, deformation, friction, and contact state, fragile household work becomes guesswork.

For buyers, this means the phrase "AI-powered" is not specific enough. Ask what the robot senses at the hand, wrist, arm, base, and body. Ask whether the company publishes slip detection, grip-force ranges, tool-use tests, spill tests, and intervention rates. A home robot does not need human-level hands on day one, but it does need enough physical awareness to avoid expensive or unsafe mistakes.

What should buyers ask before trusting a contact-rich chore demo?

Use a stricter checklist for any robot promising laundry, dishes, pet care, cooking help, plant care, or elder-assistance handling.

1. How many full task attempts were run? A single success clip is weak. A table of 100 attempts, failure categories, and recovery rates is stronger.
2. Was the task reset by a person between tries? A robot that succeeds only after a human arranges the towel, cup, tray, or litter box is not doing the real chore.
3. What contact sensors are active? Look for force/torque sensing, tactile fingertips or skin, slip detection, compliant actuation, and wrist or palm sensing.
4. Can it abort safely? The robot should know when to stop pulling a drawer, stop pouring, release a fragile object, or ask for help.
5. Does it recover from common failures? Recovery is the product. Bad grasps, partial folds, stuck tools, object slips, pet interruptions, and occluded cameras are normal at home.
6. Is remote assistance part of the workflow? Teleoperation is not cheating if it is disclosed. Weave Isaac 0 is explicit that a specialist can help if the laundry-folding robot gets stuck. That is more honest than pretending early autonomy never fails.
7. What does the warranty cover after contact damage? Contact-rich chores wear grippers, fingertips, actuators, fabric covers, and sensors. Ask about repair cost before you buy.

Are contact-rich home chores ready in 2026?

Some narrow versions are ready, but general household contact-rich autonomy is not. The honest 2026 answer is layered:

• Ready now: stationary or constrained systems for one chore, such as laundry folding with remote-assist fallback, or mobile manipulators used by researchers and assistive pilots.
• Promising but early: home humanoids with soft bodies, tactile skin, compliant hands, and expert/teleoperation modes.
• Not proven yet: a general robot that can walk into an arbitrary home and reliably fold laundry, serve drinks, scoop litter, open stuck containers, clean up spills, and recover without a human reset.

That should not make buyers cynical. It should make them specific. The next wave of useful home robots will not be defined only by smarter chat, faster walking, or better object recognition. It will be defined by the boring physical details: how hard the robot presses, whether it notices slip, how it keeps balance while using both hands, how quickly it gives up before damage, and how much evidence the company publishes beyond the best take.

For now, contact-rich chores are the filter. If a robot company can explain force, touch, recovery, and repeated trials, keep reading. If it only shows a clean demo in a perfect room, assume towels, tea, and cat litter are still harder than the video makes them look.