Home service robots (domestic mobile manipulation)

Scope and source summary

Domestic mobile manipulation is not a single architecture. It spans at least four subcategories:

1. Wheeled cleaning robots with actuated cleaning elements — the commercial backbone, led by robot vacuums/mops.
2. Mobile platforms with a robot arm or gripper — often still research or low-volume teleop platforms.
3. Home patrol/telepresence robots — mobile camera/screen systems, sometimes sold as security or companion devices.
4. Research platforms used for open-vocabulary “go find and move an object” tasks inside real homes.

The unifying technical problem is doing physical work in unstructured residential environments: moving people and pets, clutter, occlusion, variable lighting, floor transitions, children’s toys, rugs, thresholds, and furniture.

This note uses the following strongest sources:

• Granted patent: US 8,380,350 B2, “Autonomous coverage robot navigation system,” assigned to iRobot Corporation, granted 2013-02-19. This is a T1 legal-record source for navigation/coverage claims, not a performance test. [1]
• Standards-body document: ISO 13482:2014, Robots and robotic devices — Safety requirements for personal care robots. This is a T1 standards source for safety architecture and hazard identification. [2]
• Peer-reviewed paper: Argall, Chernova, Veloso, and Browning, “A survey of robot learning from demonstration,” Robotics and Autonomous Systems, 2009, DOI 10.1016/j.robot.2008.10.024. [3]
• Two arXiv preprints labeled as preprints: Mobile ALOHA and OK-Robot. Neither is used as a stronger source than the-peer-reviewed paper or granted patent. [4][5]
• Company/regulatory filings: SEC EDGAR filing indexes for iRobot and Amazon, used to cross-check adoption and risk language. [6][7]
• Reputable news for adoption/discontinuation: The Verge report on Amazon Astro for Business. [8]

Core technology stack

2.1 Mobile navigation, localization, and coverage

The dominant deployed domestic mobile robot is a wheeled coverage robot. Its navigation system must build or receive a map, localize the robot in that map, plan paths, avoid obstacles, and execute coverage behavior.

US 8,380,350 B2 describes a navigation and coverage system for a mobile robot. It is a granted patent, not an application, and its claims are legal records. The patent’s public abstract describes a control system that uses stored map data and enables the robot to perform surface coverage. [1]

The important technical distinction for a deep-dive is:

A granted patent shows what an assignee claimed as its invention. It does not show measured success rate, failure rate, battery life, or field reliability. Those performance attributes require test reports, benchmarks, or regulator data.

ISO 13482:2014 is relevant because a domestic mobile manipulator is a “personal care robot” under the standard. The standard requires a risk assessment covering hazards from locomotion, grasping, contact, and unintended motion in non-industrial environments. [2] This is a design-process requirement, not a performance certification.

2.2 Manipulation and learning from demonstration

Manipulation in homes is harder than navigation because the robot must reason about object geometry, material, fragility, and task context.

Argall et al. surveyed robot learning from demonstration (LfD). The paper points out that demonstration data can be mapped into policies, but the central technical problems are correspondence, generalization to new environments, and sample efficiency. [3] This is a peer-reviewed T1 source and supports a conservative view of autonomous home manipulation: demonstrations alone do not solve generalization.

Mobile ALOHA is an arXiv preprint. It presents a low-cost whole-body teleoperation system for bimanual mobile manipulation, with the goal of collecting demonstration data and training imitation-learning policies. The system is a research platform, not a shipping product. [4]

OK-Robot is another arXiv preprint. It studies open-vocabulary mobile manipulation by integrating pre-trained navigation and vision-language models. It is a benchmark/architecture paper, not a consumer product. [5]

Together, these sources show the field’s current approach:

SLAM + motion planning + learned manipulation policies + pre-trained vision-language models.

But the peer-reviewed grounding is thinner for the manipulation half than for the navigation half.

2.3 Safety and human-robot interaction

Domestic robots operate near children, elderly people, pets, and people who may be asleep or moving unpredictably. ISO 13482:2014 separates personal care robots into categories including mobile servant robots and requires safety-related parts of the control system to be designed under a risk-assessment process. [2]

This is one reason true home mobile manipulators with arms have not scaled: an arm adds contact hazards, pinch points, dropped-object risk, and higher controller complexity. A vacuum robot’s primary “manipulation” actuator is low-energy floor contact; a general-purpose arm is not.

Quantitative worked calculation: battery-limited runtime

A useful sanity check for domestic mobile robots is whether drive power or task power dominates battery life.

Formula:

\[

P_{traction} = \frac{\mu_{roll} \cdot m \cdot g \cdot v}{\eta}

\]

\[

t = \frac{E_{usable}}{P_{traction} + P_{payload}}

\]

Assumed inputs

Calculation

Traction power at the wheels:

\[

P_{traction} = 0.05 \times 4.0 \times 9.81 \times 0.30 = 0.5886\ \text{W}

\]

Divide by drive efficiency:

\[

P_{drive,input} = \frac{0.5886}{0.70} = 0.8409\ \text{W}

\]

Total average electrical load if the cleaning/manipulation payload draws 20 W:

\[

P_{total} = 0.8409 + 20.0 = 20.84\ \text{W}

\]

Runtime:

\[

t = \frac{26}{20.84} = 1.25\ \text{hours}

\]

Sensitivity

• If \(\mu_{roll}\) doubles to 0.10, drive input rises to about 1.68 W, and total runtime falls only to about 1.18 h.
• If robot mass doubles to 8 kg, drive input rises to about 1.68 W, and runtime falls to about 1.20 h.
• If payload load drops to 10 W, runtime rises to about 2.40 h.
• If payload load rises to 40 W, runtime falls to about 0.64 h.

Interpretation

Drive power alone is a small part of the energy budget. The dominant term is the task payload — suction/blower, water pump, brush motor, arm actuators, or compute. This is why robot vacuum runtime is tightly related to cleaning mode and why mobile manipulators with arms have severe autonomy limitations unless the arm is off or idle most of the time. This calculation is a derived sanity check, not a product performance claim. The numerical inputs are not cross-verified and must be validated against a specific robot’s datasheet.

IP and regulatory landscape

4.1 Patent evidence: US 8,380,350 B2

• Document identifier: US8380350B2
• Grant date: 2013-02-19
• Assignee: iRobot Corporation
• URL: patents.google.com
• Access date: 2026-05-07

This is a granted patent, not an application. It is a continuation of earlier iRobot coverage-navigation filings and is part of a family that has been asserted in litigation. The patent is strong evidence of a recorded navigation/coverage claim set. It is not evidence that every current product practices every claim, nor evidence of field reliability. [1]

4.2 Standards: ISO 13482:2014

ISO 13482:2014 specifies requirements and guidelines for the inherently safe design, protective measures, and information for use of personal care robots. In a domestic mobile manipulation context, this is the relevant top-level safety standard before product-specific standards. A company can claim conformance, but conformance is not the same as government approval. [2]

4.3 Peer-reviewed paper: Argall et al. 2009

This peer-reviewed survey anchors the manipulation-learning portion. It is old enough that it does not cover modern vision-language models, but it remains a T1 citation for the core problem: demonstrations are cheap to collect but hard to generalize. [3]

Adoption and company disclosure cross-check

5.1 What is commercially deployed

The commercial home robot category is dominated by robot vacuums/mops. iRobot is a pure-play specialist and its SEC EDGAR filings show the company’s consumer home robot focus. [6]

The mobile manipulation segment with arms is much smaller. Mobile ALOHA and OK-Robot are research artifacts, not consumer products. [4][5]

Amazon Astro was a domestic mobile robot with a screen/camera, not a general-purpose arm. Amazon later said it would discontinue the business version. The Verge reported on the discontinuation, which is relevant adoption context. [8]

5.2 Marketing vs filed disclosures

There is a recurring discrepancy pattern:

• Marketing language about “smart mapping” and “room-to-room navigation” implies robust operation. The patent documents show navigation claims, but the marketing does not cite a measured false-positive rate for obstacle detection or a household failure-rate dataset. [1][6]
• SEC risk-factor language is more conservative. iRobot’s 10-K filings, available through EDGAR, include risk categories such as competition, manufacturing dependence, product quality, and consumer demand. Exact quotes from a specific 10-K page were not captured in this pass; the filing index is cited, and exact language should be pulled from the accession-specific document. [6]
• Amazon’s retreat from Astro for Business is an empirical contradiction to the early enthusiasm for in-home general-purpose mobile robots. [8]

5.3 Source-bound claims

• Claim: iRobot holds a granted coverage navigation patent.

Source-bound: US8380350B2. [1]

• Claim: A safety standard exists for personal care robots including mobile servant robots.

Source-bound: ISO 13482:2014. [2]

• Claim: Learning from demonstration is a core but unsolved manipulation methodology.

Source-bound: Argall et al. 2009, DOI 10.1016/j.robot.2008.10.024. [3]

• Claim: Low-cost whole-body teleoperation can collect bimanual mobile manipulation data.

Source-bound: Mobile ALOHA arXiv:2401.02117, preprint. [4]

• Claim: Open-vocabulary navigation + manipulation can be assembled from pre-trained models.

Source-bound: OK-Robot arXiv:2401.12202, preprint. [5]

Unit economics / business model

Who monetizes this technology

1. Robot vacuum/mop vendors monetize through hardware sales, accessories, consumables, and replacement parts. iRobot’s SEC EDGAR index tracks a public pure-play in this category. [6]
2. E-commerce/consumer hardware integrators such as Amazon monetize home robots as part of a larger ecosystem. Astro’s business retreat shows that ecosystem synergy alone did not sustain a non-vacuum home mobile robot. [7][8]
3. Compute/platform suppliers monetize the silicon and toolchain used in domestic robot designs. Public examples include NVIDIA and Qualcomm. Their SEC filing indexes are primary sources. [9][10]

Margin structure

I do not reproduce specific gross-margin figures here because I did not retrieve the specific 10-K statement page in this pass. The qualitative structure is:

• Hardware margin is pressured by motors, sensors, batteries, and injection-molded parts.
• Consumables attach rate is a higher-margin recurring stream: brushes, filters, bags, mop pads.
• Service/subscription attach is emerging but not proven for home mobile manipulation.

Any hard margin number should be pulled from the relevant 10-K and marked with its accession number. If unavailable, it must be labeled “needs primary source validation.”

Who captures value

At current maturity, value concentrates in the robot vacuum hardware and consumables segment. The general-purpose arm-based home mobile manipulator segment has not demonstrated a profitable recurring business model at scale. The IP layer is real but its monetization is mostly defensive or licensing-ready, not a standalone revenue machine.

Disconfirming Evidence

The thesis that domestic mobile manipulation is a scalable commercial opportunity would be contradicted by:

1. Sustained negative unit economics in robot vacuum/mop hardware, visible in actual 10-K gross margin and operating loss trends for the category’s pure-play public companies. [6][11]
2. No repeatable autonomous manipulation success in real homes outside teleoperation or narrow demonstrations; that would show the gap between navigation and true mobile manipulation is not closing. [3][4][5]
3. Regulatory or safety barriers that make ISO 13482-style personal-care-robot requirements too costly for mass-market arm-based home robots. [2]
4. Consumer return/repair or warranty abandonment demonstrating that mapping accuracy claims do not survive contact with real homes, which would contradict the marketing implied by the granted navigation patent. [1][6]
5. Large-platform retreat beyond Amazon Astro, such as multiple public integrators exiting non-vacuum home mobile robots. [8]