Last reviewed
May 1, 2026
Sources
25 citations
Review status
Source-backed
Revision
v1 · 3,989 words
Add missing citations, update stale details, or suggest a clearer explanation.
The Human Support Robot (often abbreviated HSR) is a small, single-armed mobile manipulator built by Toyota Motor Corporation as a research platform for domestic and assistive robotics. The original goal was modest and very Japanese in flavour: build a low-cost helper that an elderly or mobility-impaired person at home could ask, by voice or tablet, to fetch a dropped pen, pick up a TV remote, or open a curtain. What started as a single 2012 prototype shown at a Tokyo home-care exhibition turned, over the next decade, into one of the most widely shared standardised research platforms in academic mobile manipulation, mainly because Toyota loaned it to dozens of universities and because the RoboCup Federation chose it as the official robot of the @Home Domestic Standard Platform League in 2017.
HSR is a roughly 1 m tall, 37 kg cylindrical mobile robot with one telescoping arm, a two-finger gripper, an omnidirectional wheeled base, and a head that carries a microphone array, an RGB-D camera, a wide-angle camera and a stereo RGB camera. It is designed to drive into a normal Japanese living room, see the person and the objects, and pick up small things from the floor or hand them to a seated user. In Toyota's own framing it sits inside the older Toyota Partner Robot family, which also produced the trumpet-, drum- and violin-playing humanoids shown at Expo 2005 in Aichi and Expo 2010 Shanghai.
The platform paper by Takashi Yamamoto, Koji Terada, Akiyoshi Ochiai, Fuminori Saito, Yoshiaki Asahara and Kazuto Murase ("Development of Human Support Robot as the research platform of a domestic mobile manipulator", ROBOMECH Journal, published 2019) is the canonical hardware reference. Most numbers in this article come from that paper, the Toyota Global press releases, the Toyota UK datasheet write-up and the independent ROBOTS guide entry.
The direct ancestor of HSR is the Toyota Welfare Vehicle programme, started inside Toyota in 1981 to develop wheelchair-accessible cars and lift-equipped vans. That programme gave Toyota three decades of contact with assistive-equipment users and rehabilitation hospitals before it ever built a humanoid. The lessons fed into the Toyota Partner Robot project announced in 2004, which Toyota positioned as a long-term answer to Japan's ageing society. Early Partner Robot outputs were demonstration humanoids: a bipedal trumpet player and other instrument-playing robots at the 2005 World Expo in Aichi, a violin-playing version with seventeen joints in the arms and hands shown in 2007, the i-Foot mountable two-legged transporter, and the wire-actuated "Version 4" robot.
The HSR prototype was the first public attempt at the home-living concept. Toyota unveiled it on 22 September 2012 and showed it from 26 to 28 September at the 39th International Home Care and Rehabilitation Exhibition at Tokyo Big Sight. The prototype already had the cylindrical body, the folding arm, the two-finger gripper, omnidirectional wheels and a head full of cameras (a Kinect-class depth sensor plus stereo RGB). Toyota had been running in-home trials with disabled users at the Foundation for Yokohama Rehabilitation Service since 2011 and had developed the design with input from the Japan Service Dog Association: the brief was, in effect, "do what a service dog does for someone with limited limb mobility, but with a suction tip and a gripper instead of a mouth."
In July 2015 Toyota launched a revised version (often referred to as HSR R1 in academic papers) and founded the HSR Developers' Community, a programme that loans the robot to selected universities in exchange for shared software and research. By the early 2020s, Toyota was reporting around 44 partner organisations across 12 countries; in Europe alone it supports labs in Germany, the United Kingdom, France and the Netherlands.
The RoboCup Federation adopted HSR as the official platform of the new Domestic Standard Platform League (DSPL) in 2017. The first competition with the standard platforms was held at RoboCup 2017 in Nagoya, Japan, where the @Home league was split into the Open Platform League (no hardware restrictions), the Domestic Standard Platform League (Toyota HSR), and the Social Standard Platform League (SoftBank Robotics Pepper, owned at the time by SoftBank). HSR was also picked as the platform for the Partner Robot Challenge (Real Space) at the World Robot Summit, Japan's METI-sponsored international competition that ran in 2018 and 2020.
While HSR continued as the standard academic platform, research interest at Toyota itself shifted toward foundation-model robotics inside Toyota Research Institute, founded in January 2016 with a $1 billion, five-year commitment and headquartered in Los Altos, California (with offices in Cambridge, Massachusetts, and Ann Arbor, Michigan). TRI's 2023 announcement of Diffusion Policy with Shuran Song's group at Columbia University, and its Punyo soft-bodied torso for whole-body manipulation, became the public face of the lab's robotics work, but HSR remained one of the platforms TRI and its partners actually run experiments on.
| Year | Event |
|---|---|
| 1981 | Toyota Welfare Vehicle programme launched, the start of Toyota's accessibility R&D |
| 2004 | Toyota Partner Robot project formally announced |
| 2005 | Trumpet- and drum-playing Partner Robots debut at World Expo in Aichi |
| 2007 | Violin-playing Partner Robot unveiled with 17 joints in arms and hands |
| 2010 | Violin-playing robot featured in the Japan Pavilion at Expo Shanghai |
| Sept 2012 | HSR prototype unveiled at the 39th International Home Care and Rehabilitation Exhibition, Tokyo Big Sight |
| 2011-2012 | In-home trials with disabled users at Foundation for Yokohama Rehabilitation Service, in cooperation with the Japan Service Dog Association |
| July 2015 | Revised HSR (R1) launched; HSR Developers' Community founded |
| Jan 2016 | Toyota Research Institute (TRI) founded; HSR becomes one of TRI's standard platforms |
| 2017 | HSR adopted as the official platform of the RoboCup@Home Domestic Standard Platform League at RoboCup Nagoya |
| 2018 | World Robot Summit Partner Robot Challenge (Real Space) held in Tokyo with HSR |
| 2019 | Yamamoto et al. publish the canonical platform paper in ROBOMECH Journal |
| 2020 | World Robot Summit 2020 Partner Robot Challenge runs again on HSR |
| 2021 | Toyota reports HSR community has grown to ~44 partners in 12 countries |
| 2023 | TRI and Columbia announce Diffusion Policy at RSS 2023; TRI introduces the Punyo soft-bodied torso |
| 2024-2026 | HSR continues as the standard DSPL platform; Toyota's robotics work expands to general-purpose humanoids with Boston Dynamics |
Headline numbers for HSR come from Toyota's own datasheet and from the Yamamoto et al. paper. The robot is intentionally short so that it can drive under a Japanese dining table and look up at a seated user. The arm is intentionally weak so that an unexpected collision does not cause injury, with the maximum payload capped at 1.2 kg in arbitrary posture. The base is omnidirectional so that the robot can sidestep around furniture in tight apartments without re-orienting its body or head camera.
| Spec | Value |
|---|---|
| Manufacturer | Toyota Motor Corporation |
| First prototype | September 2012 (Tokyo Big Sight) |
| Production research version | HSR R1, July 2015 |
| Body type | Cylindrical, telescoping torso |
| Body diameter | 430 mm (about 14.5 in) |
| Body height | 1,005-1,350 mm (telescoping) |
| Shoulder height | 340-1,030 mm |
| Weight | ~37 kg |
| Mobile base | Omnidirectional wheeled base, 3-DoF |
| Maximum speed | 0.8 km/h |
| Floor obstacles handled | Up to 5 mm bumps, slopes up to 5° |
| Total body DoF | 8-DoF body plus pan-tilt head (per Yamamoto et al.) |
| Base | 3-DoF |
| Torso (lift) | 1-DoF |
| Arm | 4-DoF lift arm |
| Head | Pan-tilt (320° pan, 120° tilt) |
| Gripper | Two-finger parallel gripper |
| Gripper opening | Up to 130-135 mm |
| Gripping force | 40 N |
| Max payload (in arbitrary posture) | 1.2 kg |
| Arm reach | Floor to 1.35 m height; ~0.45 m forward |
| Head sensors | Microphone array (4 capsules), RGB-D camera, wide-angle camera, stereo RGB camera |
| Hand sensors | Hand-mounted camera, force sensor |
| Body sensors | Inertial measurement unit, laser rangefinder (LiDAR) at the base |
| Display | Front-facing screen used for face/UI |
| Power | Removable battery + dock charger |
| Operating system | Linux + ROS-based stack |
The Yamamoto paper frames the design choices as deliberate trade-offs against a target user scenario. The 1.35 m target arm height is roughly the top shelf of a typical Japanese kitchen cabinet; the 1.2 kg payload covers the 43 object classes Toyota selected as a representative household set (cups, bottles, remote controls, books, small bags). The omnidirectional base, the suction-pad option on the gripper and the low maximum speed all flow from the same brief.
HSR runs a Linux-based onboard computer with a ROS (Robot Operating System) software stack that Toyota provides to community members as the HSRB ROS package family. The Yamamoto paper splits the software into four layers: a device-control subsystem that talks to the motors and sensors, a real-time motion-control subsystem on the robot computer, a higher-level functional subsystem made of ROS nodes (perception, navigation, manipulation, dialogue), and a user-interface subsystem that runs on the front display and any tablet client.
On top of the Toyota stack, the community has built and shared open-source tools, including MATLAB/Simulink integrations from MathWorks, ROS# bridges to Microsoft HoloLens for mixed-reality teleoperation, and competition-specific behaviour packages used by RoboCup@Home DSPL teams. The standardised API is what makes the DSPL meaningful: every team runs the same hardware and the same low-level drivers, so the score reflects perception and planning algorithms, not how good a team's mechanical engineers are. Most teams use the Navigation Stack and MoveIt for base motion and arm planning, with custom perception pipelines on top. HSR is also one of the standard platforms in graduate ROS and mobile-manipulation courses.
| Use case | Description |
|---|---|
| Domestic fetch-and-carry assistance | The original Toyota brief: bring objects to elderly or mobility-impaired users, open curtains, suction up thin items |
| Mobile manipulation benchmarks | Standardised benchmark for academic mobile-manipulation papers |
| Elderly-care research | Field studies in rehabilitation centres and care homes, particularly in Japan |
| Robot-mediated learning | Used in studies of medically homebound children attending classrooms remotely (Ahumada-Newhart et al., 2023) |
| RoboCup@Home Domestic Standard Platform League | Official competition platform since 2017 |
| World Robot Summit Partner Robot Challenge | Official platform for the Real Space track (2018, 2020) |
| Graduate teaching | Standard platform for ROS, navigation and mobile-manipulation courses |
| Foundation-model robotics | Running visuomotor imitation learning policies, including TRI's Diffusion Policy work |
| Human-robot interaction studies | Speech, gaze and gesture experiments using the head display and microphone array |
| Mixed-reality teleoperation | Bridged to HoloLens via ROS# for immersive operator interfaces |
The RoboCup@Home league was created in 2006 to push robotics teams toward useful behaviour in real domestic environments. For its first decade, teams entered with their own custom robots, which made cross-team comparison hard: a team that finished second might just have built the better robot, not the better algorithm. In 2017 the league was split into three sub-leagues. The Open Platform League kept the no-hardware-restrictions format. The Social Standard Platform League adopted SoftBank Robotics' humanoid Pepper. The Domestic Standard Platform League picked Toyota's HSR.
A DSPL competition typically includes tasks like "go to the kitchen, find the milk, bring it to the person on the sofa," "clean up a table by sorting objects into the right bins," "open the fridge, retrieve the requested item, close the fridge," and "follow this person and remember who they are." Teams are scored on how many tasks they complete autonomously and how robust their behaviour is to noise and to changes in the environment. Because every team uses the same HSR, success or failure is essentially a software story.
The DSPL has been a steady draw for international student teams since 2017. RoboCup competitions held annually at venues including Nagoya (2017), Montreal (2018), Sydney (2019), Bangkok (canceled 2020 due to COVID, replaced by virtual events), Bordeaux (2023) and beyond have all included DSPL with HSR as the platform.
HSR's role in academic robotics is more about volume than any single landmark paper. Toyota's loaner programme has put HSRs into dozens of universities, and at this point many papers per year at conferences such as ICRA, IROS, RSS and HRI use the platform. Representative work:
Toyota Research Institute (TRI) was founded in January 2016 with a $1 billion, five-year commitment from Toyota. It is headquartered in Los Altos, California, with offices in Cambridge, Massachusetts and Ann Arbor, Michigan, led from launch by Gill Pratt, formerly of DARPA. TRI's stated focus areas are autonomous driving, robotics, materials science and human-amplification AI. Its robotics group runs HSR as one of its main mobile-manipulation platforms and has published extensively on visuomotor learning.
TRI's most-discussed robotics work in 2023-2024 was the Diffusion Policy paper with Columbia and the Punyo soft-bodied torso, a tactile-skinned upper body designed for whole-body manipulation tasks like lifting boxes, moving furniture and bundling laundry. Punyo and HSR are different platforms with different goals (Punyo is a contact-rich whole-body tactile platform with no mobile base; HSR is a single-arm mobile manipulator), but they sit inside the same TRI programme. In 2024, TRI announced a partnership with Boston Dynamics to combine its foundation-model behaviour research with Boston Dynamics' new electric Atlas humanoid, signalling that the next generation of Toyota's robotics work will run on a general-purpose humanoid rather than the specialised HSR form factor, although HSR remains in active use across the academic community.
| Platform | Manufacturer | Form factor | Approximate price | DoF | Year | Focus |
|---|---|---|---|---|---|---|
| Toyota HSR | Toyota / TRI | Single-arm cylindrical mobile base, 1.0-1.35 m | Loaner only (research partner programme) | 8-DoF body + pan-tilt head | 2012 prototype, 2015 R1 | Domestic assistance, RoboCup@Home DSPL, academic research |
| TIAGo | PAL Robotics | Single-arm humanoid torso on differential base | Approximately $58,000 | Modular (typically 7-DoF arm) | 2015 | General service robotics, ROS research |
| Stretch (1, 2, 3) | Hello Robot | Telescoping single arm on a small differential base | From $17,950 (Stretch 3) | Lift + telescoping arm + wrist | 2020 | Affordable accessible mobile manipulation |
| Fetch | Fetch Robotics | Single-arm mobile manipulator | Around $100,000 (when sold) | 7-DoF arm + base | 2015 | Logistics + research; discontinued after Zebra Technologies acquisition (2022) |
| PR2 | Willow Garage | Two-arm humanoid mobile manipulator | $400,000 (historical) | 7+7 arms | 2010 | Influential research platform; Willow Garage closed 2014 |
| Boston Dynamics Spot | Boston Dynamics | Quadruped, with optional Spot Arm | $74,500 (base, 2020) | Quadruped + 6-DoF arm option | 2019 commercial | Industrial inspection; not domestic |
| Punyo | TRI | Soft-bodied torso, no mobile base | Internal research only | Custom (whole-body tactile) | 2023 | Contact-rich whole-body manipulation |
| 1X NEO Beta | 1X Technologies | Humanoid biped | Not commercially priced | Full humanoid | 2024 | Home humanoid prototype |
| Apptronik Apollo | Apptronik | Humanoid biped | Not yet broadly sold | Full humanoid | 2023 | General-purpose humanoid |
| Unitree H1 / G1 | Unitree Robotics | Humanoid biped | From ~$16,000 (G1) | Full humanoid | 2023-2024 | Lower-cost humanoid research and demo |
| Tesla Optimus | Tesla | Humanoid biped | Target ~$20-30k (announced) | Full humanoid | 2022 prototype | General-purpose humanoid |
In this comparison HSR sits closest to TIAGo and Fetch: a single-arm wheeled service robot with a head-mounted sensor stack and a research-focused price point. Stretch is the cheaper, simpler descendant of the same idea. Spot is in a different category because it is legged and aimed at industrial inspection rather than home assistance. The new wave of humanoid robots from Tesla, 1X, Apptronik, Unitree and others represents a different bet: more general but more expensive and far less safety-validated than HSR for use around frail people.
HSR is not the most capable mobile manipulator that has ever existed. Willow Garage's PR2 had two arms and far more dexterity, and TRI's own dual-arm research rigs are more flexible. What HSR did that mattered was to be standardised, available to research labs through Toyota's loaner programme, and supported with a maintained ROS stack and a competition ecosystem. That combination is rare. PR2 was discontinued when Willow Garage shut in 2014; Fetch was discontinued after Zebra Technologies acquired Fetch Robotics in 2021. HSR has now been in continuous use as an academic platform for roughly a decade, longer than any other mobile manipulator with comparable adoption.
The RoboCup@Home Domestic Standard Platform League is the cleanest example of the impact: because every team runs the same HSR, the league actually measures algorithmic progress year over year. Without a standardised platform, you cannot compare a 2018 navigation stack to a 2024 one in any honest way. The same is true of the World Robot Summit Partner Robot Challenge.
HSR also sits at the centre of Toyota's broader Partner Robot vision, which is explicitly about Japan's demographic challenge. Japan has the oldest population in the world; the country has talked openly about robotics as part of the response to its shortage of care workers. HSR is the most concrete artefact of that policy direction so far.
HSR has a single, low-payload arm. The 1.2 kg cap is enough for a bottle of soy sauce or a small kettle but not for picking up a chair or lifting a heavy laundry basket. Many tasks that humans assume "a household robot" can do are bimanual (folding sheets, carrying a tray of dishes); HSR cannot do them.
The robot is also slow. Top speed is 0.8 km/h, well below a brisk human walking pace, and the arm motion is similarly conservative. This is intentional, since safety around frail users is one of the design priorities, but it also means HSR is not a useful platform for studying fast or contact-rich manipulation. TRI's separate Punyo platform exists in part because HSR is the wrong shape and the wrong stiffness for whole-body tactile work.
The loaner programme is generous but it is still gated. HSR is not for sale to individual researchers and the partnership process selects mostly established institutions. That contrasts with Hello Robot Stretch, which is a few times cheaper than the implied research-platform cost of HSR and can be ordered directly. For a graduate student who wants their own mobile manipulator without a Toyota relationship, Stretch is the realistic option.
Finally, the new wave of humanoid robots (Apptronik Apollo, 1X NEO, Tesla Optimus, Unitree H1 and G1, Boston Dynamics electric Atlas) is starting to make HSR feel like older-generation hardware. Humanoids have more degrees of freedom, two arms, and a body that fits human-shaped environments. They are also much more expensive (other than the lower-cost Unitree platforms), much less safety-validated, and at the time of writing none of them have a mature equivalent of the RoboCup@Home DSPL ecosystem behind them. Whether the next standardised academic mobile-manipulation platform will be a humanoid or another HSR-style purpose-built robot is genuinely an open question.
HSR is still the standard platform of the RoboCup@Home Domestic Standard Platform League and is still actively used by the partner labs in Toyota's HSR Developers' Community. Toyota's Frontier Research Center continues to publish HSR-based work, including a 2024 sketch-based teleoperation interface that lets a remote operator draw on a tablet to direct the robot, and an exhibition demonstration at the World Robot Summit 2025 in Aichi. Toyota Research Institute's robotics group has been more public about its foundation-model work, including the Punyo soft torso and the partnership with Boston Dynamics on the new electric Atlas. As of 2025-2026, HSR is therefore in a curious position: it is one of the longest-running standardised mobile-manipulator research platforms still in use, but the centre of gravity of Toyota's own robotics R&D is moving toward general-purpose humanoids and toward foundation-model behaviour learning that can in principle run on whatever hardware comes next.