CtrlK
A service robot is a robot that performs useful tasks for humans or equipment, excluding industrial automation applications. The category covers machines as different as the Roomba vacuum, a Boston Dynamics Spot inspecting a refinery, a Pepper humanoid greeting bank customers, and a fleet of Starship sidewalk robots delivering burritos across a campus. What links them is not their form or price but the application: a service robot does work outside the traditional factory floor, in environments built for people.
The formal definition comes from the international vocabulary standard ISO 8373:2021, which describes a service robot as a robot in personal use or professional use that performs useful tasks for humans or equipment. The standard distinguishes service robots from industrial robots based on intended application rather than mechanical design. The International Federation of Robotics (IFR) counted more than 205,000 professional service robots sold worldwide in 2023, a 30% jump from the previous year, in its annual World Robotics: Service Robots report.
Service robots are a multi-billion dollar industry, but the category is young. The first commercially successful home service robot, the iRobot Roomba, only launched in 2002, the same year the Intuitive Surgical da Vinci received FDA clearance for general laparoscopic surgery. Most of the current fleet of restaurant servers, sidewalk delivery robots, hotel concierges and quadruped inspectors arrived in the last decade. The next decade is widely expected to be dominated by general-purpose humanoid robots from companies such as Figure, 1X, Apptronik and Tesla.
The authoritative definition is set out in ISO 8373:2021 Robotics, Vocabulary, the third edition of the international standard for robotics terminology. ISO 8373 first defines a robot as a programmed actuated mechanism with a degree of autonomy, performing locomotion, manipulation or positioning. It then divides robots by application into industrial robots, used in industrial automation such as welding and assembly on a factory line, and service robots, which perform useful tasks for humans or equipment outside industrial automation. The 2021 revision added a third top-level category, medical robot, although IFR statistics still count medical robots as a sub-class of service robots.
A service robot is therefore defined by what it does, not by what it looks like. The definition covers wheeled mobile bases, four-legged dogs, humanoids, drones, fixed manipulator arms with mobile bases, and even autonomous boats and farm tractors. ISO 8373 also requires a degree of autonomy, the ability to perform intended tasks based on current state and sensing without human intervention. Purely teleoperated machines fall outside the definition.
The most important boundary is the one with industrial robots. ISO 10218, the safety standard for industrial robots, applies to robot arms used for welding, palletising and machine tending in factories. Service robots are explicitly outside that scope. Collaborative robots (cobots) such as those from Universal Robots, Doosan and Techman are technically industrial robots under ISO 10218, but many are deployed in laboratories, hospital pharmacies and retail kitchens for service tasks. The 2025 revision of ANSI/A3 R15.06 has stopped using the word "cobot" in favour of "collaborative applications".
The IFR, working from the ISO 8373 vocabulary, splits service robots into two main classes based on the intended user.
Personal and domestic service robots are built for use by non-professionals. Anyone can buy one in a consumer electronics shop and operate it without specialist training. The category includes household vacuum and mopping robots, robotic lawn mowers, pool cleaners, window cleaning robots, social companions, educational and entertainment robots, and assistive robots for elderly care.
Professional service robots are operated by trained personnel in commercial, institutional, military or scientific settings. They include cleaning robots in airports, robots that move pallets in warehouses, food delivery robots in restaurants, surgical robots, agricultural robots, inspection robots in oil refineries, security patrol robots, defence and search-and-rescue robots, and humanoids being trialled on warehouse floors.
A single platform can sit in either category depending on how it is sold. Pepper was a domestic robot when sold to Japanese families and a professional service robot when SoftBank deployed it as a bank greeter.
The history of the service robot starts long before the term was coined. Teleoperated manipulators for handling radioactive materials appeared in nuclear research labs in the late 1940s, and powered prosthetic limbs were under development by the 1960s. Most histories of the autonomous service robot begin with Shakey the Robot, built at SRI International between 1966 and 1972 with DARPA funding. Shakey was the first general-purpose mobile robot able to reason about its actions, using a TV camera and bump sensors and a planning algorithm called STRIPS to push boxes between rooms. The project produced the A* search algorithm, the Hough transform, and the visibility graph method.
The 1980s and 1990s saw the first commercial domestic robots. Husqvarna launched the world's first robotic lawn mower, the Solar Mower, in 1995, followed by the battery powered Automower in 1998. Sony AIBO, the entertainment robotic dog, was launched in Japan on 11 May 1999 and sold more than 170,000 units, although Sony discontinued the original line in 2006. AIBO is widely seen as the spark that turned the consumer robot from a research project into a real industry.
The true commercial breakthrough came in 2002, when iRobot released the Roomba. The Roomba was cheap (about $200 in its original form), did one thing well, and found a global market; iRobot has since sold tens of millions of units. The same period saw the surgical robot move from research to mainstream use: the Intuitive Surgical da Vinci system, descended from late-1980s SRI International research, received FDA clearance for general laparoscopic surgery in 2000.
In 2005 French startup Aldebaran Robotics released the small humanoid NAO, a fixture of research labs and university classrooms. Aldebaran was acquired by SoftBank in 2012, and in June 2014 Masayoshi Son introduced Pepper, a 1.2 metre humanoid marketed as a social robot for Japanese homes and shop floors. Pepper sold around 27,000 units before production was paused in 2021. SoftBank Robotics Europe was sold to the German United Robotics Group in 2022, reverted to its original Aldebaran name, then filed for bankruptcy in February 2025; in July 2025 China's Maxvision Technology acquired the assets, including the IP for Pepper and NAO.
The 2010s and early 2020s brought an explosion of new service robot categories. Starship Technologies, founded in 2014 by Skype co-founders Janus Friis and Ahti Heinla, reached over 9 million sidewalk deliveries by late 2025. Boston Dynamics commercialised the quadruped Spot in 2020 for industrial inspection. China-based Pudu and Keenon turned the cat-faced restaurant server robot into a global sub-industry. Denmark's UVD Robots sold large numbers of UV-C disinfection robots to hospitals during the COVID-19 pandemic. By the mid 2020s the dominant story had become the humanoid wave: Figure, 1X, Apptronik Apollo, Tesla Optimus, Sanctuary Phoenix, UBTECH, Unitree and dozens of Chinese startups racing to build a general-purpose bipedal worker.
The IFR organises service robots by application class. The largest categories, with representative platforms and vendors, are listed below.
| Application area | Typical use cases | Example platforms and vendors |
|---|---|---|
| Domestic floor cleaning | Vacuuming, mopping, bin emptying | iRobot Roomba, Roborock, Ecovacs Deebot, Eufy, Dyson 360 |
| Lawn and garden | Robotic mowing, weeding | Husqvarna Automower, Worx Landroid, Mammotion Luba |
| Pool and window cleaning | Underwater and glass cleaning at home | Maytronics Dolphin, Polaris, HOBOT, Ecovacs Winbot |
| Social and companion | Greeting, entertainment, elder care | NAO, Pepper, ElliQ, Lovot, Friday |
| Education and research | AI research platforms | LEGO MINDSTORMS, NAO, TIAGo, Toyota HSR |
| Surgical and medical | Robotic surgery, pharmacy | Intuitive Surgical da Vinci, Stryker Mako, Medtronic Hugo, ROSA |
| Rehabilitation | Powered exoskeletons, gait training | ReWalk, Ekso, Cyberdyne HAL, Hocoma Lokomat |
| Telepresence | Remote presence in offices and hospitals | Beam, Double Robotics, Ohmni, InTouch RP series |
| Hospitality and hotel | Food delivery, room service, greeting | Bear Robotics Servi, Pudu BellaBot, Keenon DinerBot, Savioke Relay |
| Last-mile delivery | Sidewalk and on-road parcel delivery | Starship Technologies, Nuro, Serve Robotics, Coco, Kiwibot |
| Indoor delivery | Hospital, office and warehouse logistics | Aethon TUG, Diligent Robotics Moxi, Relay |
| Cleaning (professional) | Floor scrubbing in airports, offices, retail | SoftBank Whiz, Avidbots Neo, Brain Corp scrubbers |
| Inspection and security | Industrial monitoring, patrols | Boston Dynamics Spot, ANYbotics ANYmal, Ghost Robotics Vision-60, Knightscope K5, Cobalt |
| Agriculture | Weeding, harvesting, autonomous tractors | John Deere autonomous tractor, Bear Flag Robotics, Naio, Ecorobotix |
| Construction | Layout, drilling, surveying | Hilti Jaibot, Dusty Robotics, Built Robotics |
| Disinfection | UV-C disinfection | UVD Robots (Blue Ocean Robotics), Xenex LightStrike |
| Defence and rescue | Reconnaissance, EOD, disaster response | Endeavor PackBot, Ghost Vision-60, QinetiQ Talon, Quince |
| Humanoid (general purpose) | Manufacturing, retail, future homes | Figure 02 / 03, 1X NEO, Apptronik Apollo, Tesla Optimus, Sanctuary Phoenix, UBTECH Walker, Unitree H1 / G1 |
The single largest application class by units sold in 2023 was transportation and logistics, which alone accounted for around 113,000 of the 205,000 professional service robots sold that year. Hospitality came next at over 54,000 units, followed by agriculture with around 20,000. Medical robots are smaller in unit terms (around 6,100 in 2023, climbing to 16,700 in 2024) but generate disproportionately large revenues since surgical systems can sell for over a million dollars each.
The most widely cited source of service robot statistics is the IFR's annual World Robotics: Service Robots report. Headline numbers from the 2024 and 2025 editions are summarised below.
| Metric | 2023 | 2024 |
|---|---|---|
| Professional service robots sold worldwide | ~205,000 units (+30% YoY) | ~199,000 units (+9% in value) |
| Asia-Pacific share of professional units | 162,284 | dominant share led by China |
| Europe share of professional units | 33,918 | strong growth in cleaning and inspection |
| Americas share of professional units | 8,927 | continued growth in delivery and security |
| Largest application class | Transportation and logistics (~113,000) | Transportation and logistics |
| Hospitality robots | 54,000+ (+31%) | continued double-digit growth |
| Agricultural robots | ~20,000 (+21%) | continued growth |
| Medical robots | ~6,100 (+36%) | ~16,700 (+91%) |
| Number of supplier companies tracked | over 1,200 globally | over 1,200 globally |
Market value estimates vary depending on what is counted. Industry trackers such as Statista, Interact Analysis and Mordor Intelligence put the global service robotics market at roughly $40-50 billion in 2024 and project well over $100 billion by 2030, with the fastest-growing segment expected to be humanoid robots (around $2.9 billion in 2025, projected to exceed $15 billion by 2030).
The IFR also publishes a parallel report on consumer service robots, mostly floor cleaners, lawn mowers and pool cleaners. Consumer unit sales are an order of magnitude larger than professional sales: iRobot alone has shipped over 40 million Roombas, and the global robotic vacuum market sells tens of millions of units a year.
A modern service robot is a fusion of mobile robotics, perception, manipulation, machine learning and human-robot interaction software.
| Building block | What it does | Common implementations |
|---|---|---|
| Localisation and mapping | Building a map of the environment and finding the robot in it | Lidar SLAM, VSLAM, ORB-SLAM, Cartographer, Pharos (ANYbotics) |
| Navigation and motion planning | Choosing collision-free paths and following them | A*, RRT, MPC, Nav2, Boston Dynamics' Autowalk |
| Locomotion | Moving the body | Differential drive wheels, omnidirectional bases, four-legged gaits trained with reinforcement learning, bipedal walking |
| Manipulation | Grasping and using objects | Two-finger grippers, three-finger hands, suction, full five-finger humanoid hands |
| Perception | Recognising objects, people, signs, gauges | RGB cameras, depth cameras, lidar, thermal imaging, neural object detectors |
| Speech and language | Understanding commands, talking back | ASR, NLU, LLM-based dialogue, on-device wake-word detection |
| Planning and reasoning | Deciding what to do next | Behavior trees, finite state machines, vision-language-action models such as RT-2, OpenVLA, NVIDIA Isaac GR00T |
| Cloud and fleet management | Telemetry, OTA updates, fleet learning | Boston Dynamics Orbit, AWS RoboMaker, custom cloud stacks |
| Safety systems | Stopping the robot before it hurts someone | Bumper sensors, safety-rated lidar, force/torque limiters, emergency stops, ISO 13482 compliance |
| Power and charging | Keeping the robot running | Lithium-ion batteries, autonomous docking, in many cases hot-swappable packs |
| Human-robot interaction | Talking, gesturing, reading intent | Touchscreens, faces, lights, sound, body posture, gesture recognition |
Since 2023 the field has moved from hand-engineered behaviour stacks toward end-to-end learned models. Google DeepMind's RT-2, the open-source OpenVLA from Stanford, and NVIDIA's Isaac GR00T family are vision-language-action (VLA) models that take a camera image and a natural-language instruction and directly output robot actions. They underpin the bet that the new generation of humanoid service robots can be trained once and then deployed on many different tasks.
Most service robots are built on top of the open-source Robot Operating System (ROS), originally developed at Willow Garage in 2007 and now maintained by the Open Source Robotics Foundation. ROS 2 is the de facto framework for most non-humanoid service robots.
The distinction between service, industrial and collaborative robots is sometimes blurred in marketing language, but the international standards still draw clear lines.
| Property | Industrial robot | Collaborative robot (cobot) | Service robot |
|---|---|---|---|
| Primary standard | ISO 10218 / ANSI R15.06 | ISO 10218 + ISO/TS 15066 (collaborative applications) | ISO 13482 + ISO 8373 |
| Typical environment | Caged area on a factory floor | Shared workspace in a factory or lab | Human environment: home, office, hospital, sidewalk, field |
| Typical user | Trained line operator, automation engineer | Trained operator, often working alongside the robot | Anyone (consumer) or trained operator (professional) |
| Mounting | Usually fixed to the floor or a gantry | Often fixed but lighter | Usually mobile |
| Typical payload | High (tens to hundreds of kg) | Low to medium (3-25 kg) | Varies; mostly low |
| Speed limits | High; safeguarded by enclosure | Reduced when humans are nearby | Speed limited for safety in human environments |
| Example platforms | KUKA KR series, Fanuc M-series, ABB IRB | Universal Robots UR series, Doosan M, Techman, Kuka iiwa | Roomba, Pepper, Spot, Starship robots, Pudu BellaBot |
The IFR uses a simple rule of thumb: an industrial robot does industrial work, a service robot does everything else, and cobots are a flavour of industrial robot that can share space with humans.
The service robot industry is more fragmented than the industrial robot business, which is dominated by the "Big Four" of Fanuc, KUKA, ABB and Yaskawa. The IFR tracks more than 1,200 service robot suppliers worldwide; some of the most prominent are listed below.
| Company | Headquarters | Main products | Notes |
|---|---|---|---|
| iRobot | United States | Roomba, Braava | Pioneer of consumer service robots; over 40 million units sold |
| Roborock, Ecovacs | China | Robotic vacuums, mops, window and lawn robots | Lead the global premium home cleaning market |
| Husqvarna | Sweden | Automower lawn mowers | Created the robotic mower category in 1995 |
| Maytronics | Israel | Dolphin pool cleaners | Largest pool cleaning robot brand worldwide |
| Aldebaran (formerly SoftBank Robotics Europe) | France | NAO, Pepper | Sold to United Robotics Group 2022; IP bought by Maxvision in 2025 |
| PAL Robotics | Spain | TIAGo, REEM, ARI, Kangaroo | Major European service and humanoid platform supplier |
| Toyota Research Institute | Japan / United States | Human Support Robot (HSR) | Open research platform used in RoboCup@Home |
| Boston Dynamics | United States | Spot, Stretch, Atlas | Owned by Hyundai; over 1,500 Spots in service |
| ANYbotics | Switzerland | ANYmal quadruped inspection robot | Spin-out of ETH Zurich; over 200 robots deployed |
| Bear Robotics, Pudu, Keenon | United States, China | Restaurant and hotel servers | Three of the largest hospitality robot vendors |
| Knightscope, Cobalt | United States | Security patrol robots | K5 outdoor and indoor variants |
| Starship, Nuro, Serve | Estonia / United States | Last-mile delivery robots | Sidewalk and on-road autonomous delivery |
| Diligent Robotics, Aethon, Relay (Savioke) | United States | Indoor delivery for hospitals and hotels | Mobile manipulators and tray robots |
| Avidbots, Brain Corp | Canada / United States | Commercial floor cleaners | BrainOS powers many third-party machines |
| Blue Ocean Robotics | Denmark | UVD disinfection, GoBe telepresence | Conglomerate of service robot brands |
| UBTECH, Unitree | China | Humanoid and quadruped robots | Listed humanoid maker and disruptive low-cost vendor |
| Figure, 1X, Apptronik, Sanctuary, Tesla | United States, Norway, Canada | General-purpose humanoid service robots | Apptronik valued at $5 billion in 2026; Tesla unveiled Optimus Gen 3 in October 2025 |
Service robots are covered by an overlapping patchwork of international, national and industry standards.
| Standard or regulation | Scope | Notes |
|---|---|---|
| ISO 8373:2021 | Robotics vocabulary | Defines robot, service robot, industrial robot and medical robot |
| ISO 13482:2014 | Safety requirements for personal care robots | Covers mobile servant, physical assistant and person-carrier robots; first ISO safety standard for service robots in non-medical settings |
| ISO 10218 | Industrial robot safety | Applies to robot arms used in industrial automation; technically not a service robot standard but often referenced |
| ISO/TS 15066:2016 | Collaborative robot applications | Detailed force and pressure limits for human contact; integrated into ISO 10218-2:2025 |
| ANSI/A3 R15.06-2025 | United States adoption of ISO 10218 | The 2025 revision is the most significant update to U.S. industrial and collaborative robot safety in more than a decade |
| IEC 60601 family | Medical electrical equipment | Applies to robots used as medical devices, including surgical robots |
| EU Machinery Regulation 2023/1230 | Machinery placed on the EU market | Replaces the older Machinery Directive; explicitly addresses AI and autonomy |
| GDPR | Data protection in the European Union | Applies to any service robot that captures personal data, including video in homes and offices |
| FAA Part 135 / Part 137 | United States aviation rules | Govern delivery drones operating in the U.S. national airspace system |
| Local sidewalk and street rules | Right of way for delivery robots | Many U.S. states and European cities now have specific delivery robot statutes |
ISO 13482 is the standard most often cited for consumer service robots. It identifies three types of personal care robot, mobile servant, physical assistant and person carrier, and lays out the main hazards (collision, sharp edges, electrical, energy storage, software faults) that designers must address. It does not cover toys, medical devices or industrial machines.
Regulation is evolving fastest in the delivery and humanoid spaces. Several U.S. states now have laws specifically governing personal delivery devices. The EU's AI Act, in force from 2024, will apply additional risk-based obligations to robots incorporating high-risk AI systems.
A few clear trends are visible in the 2024-2026 service robotics market.
The humanoid wave. The clearest single shift is the bet that a general-purpose bipedal robot will eventually outperform purpose-built service robots on a wide range of tasks. Figure, 1X, Apptronik, Tesla, Sanctuary, Agility Robotics, and a long list of Chinese firms including Unitree, UBTECH, Fourier Intelligence and AgiBot are all racing to build humanoids that can be deployed in warehouses, retail and eventually homes. Apptronik raised $520 million at a $5 billion valuation in February 2026 to scale Apollo, and Tesla unveiled Optimus Gen 3 in October 2025.
Foundation models in robotics. RT-2, OpenVLA, NVIDIA Isaac GR00T and π0 from Physical Intelligence are vision-language-action models that replace stacks of hand-coded behaviours with single end-to-end neural networks. Boston Dynamics has integrated Google DeepMind's Gemini Robotics models into Spot, allowing the robot to read complex gauges, identify spills and reason about what it sees on an inspection round.
China dominance in cleaning and hospitality. China is now both the largest single buyer of service robots and the largest single producer.
Cloud robotics. Manufacturers run their fleets through cloud platforms (Boston Dynamics Orbit, AWS RoboMaker, custom Pudu and Keenon stacks) that collect telemetry, push over-the-air updates and share learning across the fleet.
Privacy concerns. A robot vacuum maps your home, a hotel robot films its corridors, a security robot recognises faces. Several high-profile incidents (including a 2022 Roomba mapping image leak) have made regulators much more sensitive to what service robots see.
Convergence with cobots. As collaborative arms get cheaper and service robots get more capable manipulators, the line between a wheeled cobot and a mobile manipulator service robot is increasingly blurry. The 2025 ANSI R15.06 update dropped the word "cobot" entirely in favour of "collaborative applications".
Despite the headline numbers, service robotics still has serious unsolved problems.
Cost. A Spot robot costs around $75,000 to $150,000 depending on configuration, a TIAGo with arms is in the same range, and a humanoid such as Apptronik Apollo or Figure 02 is well above six figures. Roomba has shown a service robot can be cheap, but professional robots still carry industrial price tags.
Battery life. Most legged and humanoid service robots run for one to four hours per charge. This is enough for inspection rounds but not for an eight-hour shift.
Real-world generalisation. A demo video of a humanoid folding laundry is very different from a robot that reliably folds laundry in any of a thousand different homes. Foundation models such as Isaac GR00T and π0 target this gap, but it is still the central technical problem of the industry.
Trust and acceptance. Pepper sold only around 27,000 units before production was paused in 2021, partly because customers did not know what to do with it. Hospitals have found that nurses sometimes resist mobile robots that take corridor space without obvious benefit.
Liability. When a Roomba shreds a sock the customer absorbs the loss; when a delivery robot rolls over a pedestrian's foot the question of who is liable is much harder. Insurance and case law are still catching up.
Regulatory fragmentation. A delivery robot may be legal in Texas, restricted in California, banned in San Francisco, and require special permits in many European cities. This patchwork is one of the larger barriers to scaling on-street service robots.