# Wheeled-legged robot

> Source: https://aiwiki.ai/wiki/wheeled_legged_robot
> Updated: 2026-07-22
> Categories: Humanoid Robots, Robotics
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution to "AI Wiki (aiwiki.ai)".

A **wheeled-legged robot** (also called a wheel-legged or hybrid wheeled-legged robot) is a mobile robot that combines wheels with articulated legs so it can roll efficiently across flat, continuous ground while still stepping, climbing, or balancing over the discontinuities that stop a purely wheeled machine: curbs, stairs, gaps, and rubble. The design goal is usually summarised as "the best of both worlds", the low energy cost and high speed of wheels where the ground allows, and the terrain-negotiating ability of legs where it does not [1][2]. Most wheeled-legged robots place a driven wheel at the end of each articulated leg, but the class also spans two-wheeled balancing machines with a legged or humanoid upper body, limbs that transform between wheel and foot modes, and planetary "wheel-on-limb" rovers. As a form of [robot locomotion](/wiki/robot_locomotion), the approach sits between pure rolling and pure walking, trading some mechanical simplicity for a much wider envelope of usable terrain [3].

## Taxonomy of wheeled-legged designs

Wheeled-legged robots are usually classified by how the wheels attach to the body, by whether the wheels are actively driven or passive, and by how the machine keeps its balance. Academic surveys divide them further by leg count (bipedal, quadrupedal, hexapedal) and by "skate" configuration (one wheel per leg, inline skates, or quad skates) [3][4].

| Sub-type | Wheel drive | Balance | Representative examples | Distinguishing feature |
|---|---|---|---|---|
| Wheels-on-legs quadruped | Actively driven | Statically or dynamically stable on four contacts | [ANYmal](/wiki/anybotics) on Wheels, [Unitree](/wiki/unitree) B2-W, [DEEP Robotics](/wiki/deep_robotics) Lynx, LimX W1 | A wheel at each foot of a four-legged base; the most common industrial form |
| Wheeled-legged biped (two-wheeled balancer) | Actively driven | Dynamically balanced, inverted pendulum | [Boston Dynamics](/wiki/boston_dynamics) Handle, Ascento | Balances on two wheels while legs crouch, extend, or jump |
| Wheeled humanoid | Actively driven | Dynamically balanced on a two-wheel base | 1X EVE (Halodi) | A human-like torso and arms on a self-balancing wheeled base |
| Wheel-leg transformer | Switchable | Varies | Roller-Walker | Limbs mechanically switch between wheel mode and foot mode |
| Passive-wheel skater | Passive (unactuated) | Varies | Roller-Walker (skate mode), research skaters | Legs propel the body while free-rolling wheels reduce friction, like a skater |
| Wheel-on-limb rover | Actively driven | Statically stable, high [degrees of freedom](/wiki/degrees_of_freedom) | NASA JPL RoboSimian, ATHLETE | High-DoF limbs tipped with wheels for planetary terrain |

The active-versus-passive split is the most consequential engineering choice. **Actively driven wheels** carry their own motors and can propel, brake, and even stand the robot up, enabling high-speed rolling, stepping, and jumping, but they add mechanical complexity, power draw, and weight to every limb [3]. **Passive wheels** are unactuated: the legs generate motion and the free-rolling wheels simply cut friction, "similar to a human on a skateboard or rollerblades" [3]. Passive designs are lighter and simpler and can be strikingly efficient, but they need a propulsion scheme (angled push legs, claws, or transformable feet) and are harder to steer and brake precisely [3]. Handle, ANYmal on Wheels, and the recent Chinese quadrupeds all use actively driven wheels; the pioneering Roller-Walker demonstrated the passive-skating idea.

## Why hybrid: the efficiency case

The rationale for adding wheels to a legged robot is grounded in energy. The standard measure is the [cost of transport](/wiki/cost_of_transport) (COT), the dimensionless energy required to move a unit of weight a unit of distance; lower is better. The comparison across locomotion modes traces back to the 1950 "specific resistance" framework of Gabrielli and von Karman, who plotted the power required to propel vehicles against their speed and showed how sharply the modes differ [5]. Walking is expensive: a legged robot must repeatedly accelerate and decelerate its limbs and support its body against gravity at every step. Rolling on wheels avoids most of that, so on smooth, level ground wheels have a far lower COT than legs.

The ANYmal on Wheels experiments quantify the gap on a single machine. In the ETH Zurich study "Keep Rollin'", the wheeled quadruped, "on flat terrain, the robot achieves a COT of 0.1 while driving 2 m/s and the mechanical power consumption is 63.64 W" [1]. Measured on the same robot, that rolling COT was "lower by 83% w.r.t. the trotting gait and by 17% w.r.t. skating motions" with passive wheels [1]. Driving also raised the top speed: "with 4 m/s we broke ANYmal's maximum speed record of 1.5 m/s" set by pure legged locomotion [1]. In other words, switching from stepping to rolling on flat ground cut the energy per metre by more than fivefold and roughly tripled the achievable speed on the identical platform.

Legs, in this framing, are not the everyday mode; they are the exception mode, deployed only to cross the discontinuities that wheels cannot. A robot that rolls 95% of the time and steps for the remaining 5% spends most of its energy budget at the low COT of a wheel and pays the high COT of a leg only briefly. Surveys of passive-wheel designs report COT reductions of 70% to 80% versus walking for skating locomotion, reinforcing the same conclusion from a different mechanical starting point [3]. This efficiency argument, more than raw speed, is why the form factor keeps reappearing in logistics and inspection, where battery life and range decide whether a deployment is economic.

## History and key platforms

Wheeled-legged robots have a long research lineage that turned into commercial products only in the 2020s. The timeline below lists representative platforms; the paragraphs that follow add sourced detail.

| Platform | Developer | Year | Description |
|---|---|---|---|
| Roller-Walker | Endo and Hirose, Tokyo Institute of Technology | 1999 | Quadruped whose feet flip into passive wheels for skating, an early efficiency demonstration [3] |
| ATHLETE | NASA Jet Propulsion Laboratory | 2000s | Six-limbed wheel-on-limb freight rover for planetary operations [22] |
| RoboSimian | NASA JPL | 2015 | Four-limbed wheel-on-limb robot, fifth place at the DARPA Robotics Challenge Finals [22] |
| Handle | [Boston Dynamics](/wiki/boston_dynamics) | 2017 | Two-legged wheeled robot balancing on two wheels; 2019 logistics version built pallets [10][11][12] |
| ANYmal on Wheels | ETH Zurich Robotic Systems Lab | 2018 to 2019 | Research wheeled [quadruped](/wiki/quadruped_robot); source of the COT and speed results [1][2] |
| Ascento | ETH Zurich spin-off | 2017 to 2023 | Two-wheeled balancing legged robot; Ascento Guard for security patrol [13][14][15] |
| Swiss-Mile / RIVR | ETH Zurich Robotic Systems Lab spin-off | 2023 | Wheeled-legged ANYmal derivative for delivery; acquired by Amazon in 2026 [6][7][8] |
| B2-W | [Unitree](/wiki/unitree) | 2024 | Wheeled version of the B2 industrial quadruped [18] |
| Lynx (M20) | [DEEP Robotics](/wiki/deep_robotics) | 2024 to 2025 | Industrial-grade wheel-legged quadruped for inspection [17] |
| W1 | LimX Dynamics | 2024 | Perceptive wheeled quadruped that also balances and walks on two legs [19][20] |
| EVE | 1X (formerly Halodi Robotics) | 2020 | Wheeled self-balancing humanoid for logistics and security [21] |

### Early lineage: skaters and planetary rovers

The idea of a leg that ends in a wheel is decades old. Shigeo Hirose's Roller-Walker (from 1999) let a quadruped flip its feet into passive wheels and skate, showing early that free-rolling wheels could sharply cut the energy of legged travel [3]. In parallel, NASA JPL pursued "wheel-on-limb" designs for space: the six-legged ATHLETE freight rover and, later, RoboSimian, a four-limbed robot with 28 actuated joints and wheels on its limbs that placed fifth at the 2015 DARPA Robotics Challenge Finals [22]. Notably, four of the five best-placed DRC teams combined legged with wheeled locomotion, an early real-world signal that hybrids outperform either mode alone on mixed terrain [22]. RoboSimian survives today as an actively articulated wheel-on-limb rover for unstructured planetary analogue terrain.

### ETH Zurich, ANYmal on Wheels, and the RIVR spin-out

The modern wheeled-legged quadruped was largely defined at ETH Zurich's Robotic Systems Lab (RSL), led by Marco Hutter. The lab's ANYmal quadruped was fitted with driven wheels at each foot, and a series of papers by Marko Bjelonic and colleagues produced both the control methods and the efficiency results cited above [1][2]. It is important not to conflate the lab's two spin-offs. **ANYbotics**, founded in 2016 by Peter Fankhauser, Marco Hutter, and Hannes Sommer, commercialises the purely legged [ANYmal](/wiki/anybotics) for autonomous industrial inspection in oil and gas, chemicals, mining, and energy; ANYmal has no wheels [24]. The wheeled-legged line became a separate company: **Swiss-Mile**, spun out of RSL in 2023 and led by Marko Bjelonic, which built a wheeled-legged robot that can autonomously switch between driving and stepping, stand and balance on its two rear wheels, open doors, and manipulate packages while upright [6][7].

Swiss-Mile rebranded to **RIVR** on 30 January 2025 [7]. In August 2024 it had raised a seed round of about 22 million US dollars co-led by Jeff Bezos through Bezos Expeditions and the venture firm HongShan, with participation from the Amazon Industrial Innovation Fund [7][8]. RIVR ran field trials for last-100-yard parcel delivery, including a collaboration with the UK carrier Evri and a Swiss Post pilot [23]. On 19 March 2026, Amazon acquired RIVR to test wheeled-legged robots for doorstep delivery; the companies did not disclose financial terms [8][9]. The RIVR robots roll on wheels between stops and use their legs to handle kerbs and stairs at the door, the same rationale ETH's early experiments had quantified.

### Boston Dynamics Handle

Boston Dynamics introduced **Handle** in February 2017 as, in the company's words, "a research venture into wheeled robots" that balanced dynamically on two wheels [10][12]. Each of Handle's two legs ended in a wheel, making it effectively a wheeled-legged biped: it could crouch, roll at roughly 15 km/h, jump about 1.2 metres, and lift loads up to 50 kg [12]. In 2019 Boston Dynamics reframed Handle as "a mobile manipulation robot designed for logistics", adding a vacuum-gripper arm so it could depalletize and build mixed-SKU pallets autonomously [11]. The company later shifted its warehouse effort to **Stretch**, a robotic arm on a compact omnidirectional wheeled base rather than a balancing wheeled-legged machine; Boston Dynamics says Handle "led to the development of Stretch" [10]. Stretch traded Handle's dynamic agility for a simpler, more stable base better suited to repetitive box moving, an instructive case of a wheeled-legged research platform giving way to a plainer wheeled product for a specific job.

### Ascento

**Ascento** began as an ETH Zurich student focus project (2017 to 2018) by Alessandro Morra, Miguel de la Iglesia Valls, Dominik Mannhart, and Ciro Salzmann, who built a two-wheeled balancing robot with short articulated legs that could roll on flat ground and jump up steps [15]. The team spun the project into a company and, in 2023, launched the **Ascento Guard**, an autonomous outdoor security-patrol robot, alongside a 4.3 million US dollar funding round [13][14]. The Guard balances on two wheel-legs, patrols at roughly 4.5 km/h, weighs about 30 kg, and carries a 360-degree camera plus thermal and infrared imaging to spot intruders, fires, or floods; the later Guard 2.0 added a larger battery for up to eight hours per shift [13][15][16]. Ascento offers the robot as a rented service rather than a one-off sale.

### Chinese wheeled-legged quadrupeds

Chinese robotics firms brought wheeled-legged quadrupeds to volume in 2024. Unitree's **B2-W** is a wheeled version of its B2 industrial quadruped, with a pneumatic wheel at each foot; vendor listings cite rolling speeds of roughly 20 km/h, a moving payload around 40 kg, and long range, and Unitree's demonstrations showed it climbing, carrying a rider, and flipping [18]. DEEP Robotics' **Lynx** (the Lynx M20) is pitched as an industrial-grade wheel-legged quadruped for inspection: about 33 kg, a 15 kg operational payload, roughly 5 m/s in lab tests (2 m/s in operation), continuous stair climbing, IP66 sealing, and dual LiDAR for mapping [17]. LimX Dynamics launched **W1**, which it describes as China's first wheeled quadruped with autonomous terrain perception able to climb and descend stairs through real-time gait planning; it has 16 actuated degrees of freedom and can even rear up to balance and walk on its two hind wheel-legs [19][20].

### Wheeled humanoids

A distinct branch replaces the legs entirely below the waist with a self-balancing two-wheeled base, giving a **wheeled humanoid**: a human-like torso and arms that rolls rather than steps. The clearest commercial example is **EVE** from 1X Technologies (formerly Halodi Robotics), a self-balancing humanoid used for logistics and security whose wheeled base derives from the classic Segway-style wheeled inverted pendulum [21]. Wheeled bases are attractive for a [humanoid](/wiki/humanoid_robot) because they are cheaper, more stable, and more energy-efficient indoors than legs, at the cost of stairs. Some makers ship wheeled-base variants for exactly this reason, while others treat wheels as a stepping stone: 1X itself moved from the wheeled EVE toward the fully bipedal [NEO](/wiki/1x_neo) built by [1X Technologies](/wiki/1x_technologies). Handle can be read as the bridge between the two branches, a wheeled biped that is part balancing humanoid and part wheels-on-legs quadruped logic applied to two limbs.

## How they work

A wheeled-legged robot has to decide, continuously, when to roll and when to step, and then execute both smoothly on the same body. Three technical problems dominate.

**Hybrid locomotion planning.** The controller must choose a locomotion mode for the terrain ahead and blend transitions between them. Early ETH work solved this with model-based optimization: a zero-moment-point motion optimization feeding a hierarchical [whole-body control](/wiki/whole_body_control)ler that issues torque commands while respecting the nonholonomic rolling constraint of the wheels [1]. A follow-up decomposed the high-dimensional problem into wheel and base trajectories so it could be solved in real time on board in a model-predictive-control fashion, and deployed the robot at the DARPA Subterranean Challenge to map and explore underground [2]. Since then the field has shifted markedly from model-based controllers toward [reinforcement learning](/wiki/reinforcement_learning), which handles the underactuated dynamics and terrain variability of these machines more robustly and now drives the standing, parkour, and door-opening demos seen from RSL and its spin-outs [3].

**Balancing on driven wheels.** Two-wheeled machines such as Handle, Ascento, and EVE are inverted pendulums: they are statically unstable and must actively drive their wheels to stay upright, the same control principle as a Segway. Torque-controlled wheels let the robot modulate ground reaction forces precisely, which is what allows a quadruped like ANYmal on Wheels to rear up onto two wheels and still balance [6]. Four-wheeled quadrupeds are easier to balance but must still coordinate wheel torques with leg posture during transitions, for instance shifting from a rolling stance to a stepping gait at a stair.

**Sensing, driven versus passive wheels, and the actuator penalty.** Negotiating steps and curbs requires terrain perception, typically [LiDAR](/wiki/lidar) and depth cameras feeding a local elevation map, plus [motion planning](/wiki/motion_planning) that knows the geometry of both wheel and foot contacts. The wheel design is a core tradeoff: actively driven wheels add an [actuator](/wiki/actuator), its power electronics, and mass at the end of every leg, which raises inertia and cost but enables the fast, efficient rolling that motivates the whole design; passive wheels avoid that penalty but give up direct speed and braking control [3]. Every wheeled-legged robot is a negotiation between these extremes, and much of the engineering is about hiding the seams so the machine feels like one coherent system rather than a wheeled robot bolted to a legged one.

## Applications

The economics favour environments that are mostly flat with occasional vertical breaks, which is why warehouses, factories, and campuses are the sweet spot. Concrete applications include:

- **Industrial inspection.** Rolling efficiently between inspection points and stepping over cable trays, sills, or a short stair, with LiDAR mapping and thermal cameras, is the pitch for DEEP Robotics' Lynx and comparable quadrupeds [17]. The wheels extend battery range enough to make routine autonomous patrols economic.
- **Logistics and last-100-metre delivery.** RIVR's robots roll along pavements and driveways, then climb the steps to a front door to complete "last-100-yard" parcel drops, a task pure wheeled robots cannot finish and pure walkers do too slowly and expensively; Amazon's 2026 acquisition of RIVR was aimed squarely at this doorstep-delivery problem [8][9][23].
- **Security patrol.** Ascento's Guard patrols large outdoor sites at a walking pace for a full shift, balancing on two wheel-legs to keep sensors at a useful height while covering ground efficiently [13][16].
- **Warehouse material handling.** Handle's logistics incarnation targeted pallet building and truck unloading before Boston Dynamics moved that work to the wheeled Stretch [11].

Against a pure [autonomous mobile robot](/wiki/autonomous_mobile_robot) (AMR), the wheeled-legged machine trades some cost and reliability for the ability to leave paved, level floors; against a pure walker it trades some rough-terrain ability for range and speed.

## Comparison with other form factors

The wheeled-legged robot occupies a middle band of the [robot form factor](/wiki/robot_form_factor) spectrum, which is exactly why it is chosen only when the terrain profile justifies the added complexity.

| Attribute | Wheeled-legged robot | Wheeled AMR | Pure quadruped | Biped / humanoid |
|---|---|---|---|---|
| Flat-ground efficiency (COT) | Very low while rolling [1] | Lowest | High (legged) | High (legged) |
| Top speed on flat ground | High (roughly 4 m/s and up) [1] | High | Moderate | Low to moderate |
| Curbs and short steps | Good | Poor to none | Excellent | Excellent |
| Stairs | Limited to moderate | None | Excellent | Good, with [bipedal locomotion](/wiki/bipedal_locomotion) |
| Very rough or soft terrain | Moderate | Poor | Excellent | Moderate |
| Mechanical complexity and cost | High | Low | High | Highest |
| Payload for given mass | Moderate | High | Moderate | Low to moderate |
| Best fit | Flat sites with occasional steps | Warehouses, smooth floors | Rough outdoor terrain | Human-built spaces, manipulation |

The table makes the niche explicit: if a site is genuinely flat, a wheeled AMR wins on cost and reliability; if it is genuinely rough or stair-dominated, a pure walker wins on capability. The wheeled-legged robot is the right answer specifically for the large middle, mostly rollable ground with periodic discontinuities, where paying for both wheels and legs buys more usable range than either alone.

## Limitations

The hybrid design carries real penalties. Putting a driven wheel and its actuator at the end of every leg adds mass, cost, and failure modes to a machine that already has the full actuator count of a legged robot, so a wheeled-legged quadruped is typically more expensive and more complex than either a comparable AMR or a comparable pure quadruped [3]. The extra distal mass raises limb inertia, which can hurt the very agility the legs are meant to provide. Payload for a given total mass tends to be lower than a wheeled cart's because structure and energy are spent on articulation. On stairs and broken ground the wheels are dead weight and the robot must fall back on stepping, where a purpose-built walker is usually more capable and more sure-footed; wheeled-legged robots climb stairs, but generally more slowly and within tighter limits than the best pure legged machines [17][19]. Battery life, though better than a walker's thanks to rolling, still constrains delivery and patrol duty cycles, and the control stack that fuses rolling, stepping, and balancing is harder to make robust than either mode on its own. For many buyers the honest question is whether their terrain really needs legs at all; where it does not, a simpler robot wins.

## Outlook

Two trends are pushing wheeled-legged robots from lab demos toward products. First, control has matured: the move from hand-tuned model-based planners to reinforcement-learning policies has made standing, stair-climbing, and door-opening reliable enough to show to customers, and the same policies transfer across the growing fleet of similar quadrupeds [3]. Second, the commercial signal is now unambiguous. Amazon's 2026 acquisition of RIVR put a major logistics buyer behind wheeled-legged doorstep delivery, while Unitree, DEEP Robotics, and LimX Dynamics have brought wheeled quadrupeds to volume pricing [8][17][18][19]. The likely trajectory is not a general-purpose robot but a set of well-fitted niches: inspection on mixed industrial sites, security patrol on campuses, and last-100-metre delivery in suburbs, all places where the ground is mostly rollable but not entirely so. Whether the wheeled humanoid persists or is absorbed by cheaper bipeds remains open; 1X's shift from EVE toward a walking design suggests that, for human spaces, legs may eventually win, while for goods movement over pavement the wheels are likely to stay.

## ELI5

Wheels are great on smooth floors: they roll fast and barely use any energy. Legs are great on stairs and bumpy ground, but walking uses a lot of energy and is slow. A wheeled-legged robot cheats by having both. It puts a wheel at the end of each leg, so most of the time it just rolls along like a skateboard, which is easy and efficient. When it reaches a curb or a staircase, it stops rolling and uses its legs to step up, then goes back to rolling on the other side. On one real robot, rolling instead of walking used about five times less energy and let it go about three times faster on flat ground [1]. The catch is that adding a motor and wheel to every leg makes the robot heavier, pricier, and trickier to control, so engineers only build one when the ground is mostly smooth but has a few steps in the way.

## See also

- [Robot form factor](/wiki/robot_form_factor)
- [Cost of transport](/wiki/cost_of_transport)
- [Quadruped robot](/wiki/quadruped_robot)
- [Bipedal locomotion](/wiki/bipedal_locomotion)
- [Robot locomotion](/wiki/robot_locomotion)
- [Whole-body control](/wiki/whole_body_control)
- [Autonomous mobile robot](/wiki/autonomous_mobile_robot)
- [Humanoid robot](/wiki/humanoid_robot)

## References

1. Bjelonic, M., Bellicoso, C. D., de Viragh, Y., Sako, D., Tresoldi, F. D., Jenelten, F., Hutter, M. "Keep Rollin' - Whole-Body Motion Control and Planning for Wheeled Quadrupedal Robots." IEEE Robotics and Automation Letters, 2019. arXiv:1809.03557. https://arxiv.org/abs/1809.03557
2. Bjelonic, M., Sankar, P. K., Bellicoso, C. D., Vallery, H., Hutter, M. "Rolling in the Deep - Hybrid Locomotion for Wheeled-Legged Robots Using Online Trajectory Optimization." IEEE Robotics and Automation Letters, 2020. arXiv:1909.07193. https://arxiv.org/abs/1909.07193
3. "Passive wheels on legged robots: a survey." Frontiers in Robotics and AI, 2026. https://www.frontiersin.org/journals/robotics-and-ai/articles/10.3389/frobt.2026.1857985/full
4. "A Survey of Wheeled-Legged Robots." CLAWAR 2022 proceedings, Springer, 2022. https://link.springer.com/chapter/10.1007/978-3-031-15226-9_11
5. Gabrielli, G., von Karman, Th. "What price speed? Specific power required for propulsion of vehicles." Mechanical Engineering, 72(10), pp. 775-781, 1950. Summarised at the von Karman-Gabrielli diagram entry. https://en.wikipedia.org/wiki/Von_K%C3%A1rm%C3%A1n%E2%80%93Gabrielli_diagram
6. RIVR. "From Research to Global Impact: The Journey from Swiss-Mile to RIVR." rivr.ai. https://www.rivr.ai/stories/from-research-to-global-impact-the-journey-from-swiss-mile-to-rivr
7. The Robot Report. "Swiss-Mile rebrands to RIVR, continues developing wheeled quadrupeds." 30 January 2025. https://www.therobotreport.com/swiss-mile-rebrands-to-rivr-continues-developing-wheeled-quadrupeds/
8. TechCrunch. "Amazon acquires Rivr, maker of a stair-climbing delivery robot." 19 March 2026. https://techcrunch.com/2026/03/19/amazon-acquires-rivr-maker-of-a-stair-climbing-delivery-robot/
9. CNBC. "Amazon acquires startup Rivr to test robots for 'doorstep delivery'." 19 March 2026. https://www.cnbc.com/2026/03/19/amazon-acquires-startup-rivr-to-test-robots-for-doorstep-delivery.html
10. Boston Dynamics. "History and Legacy." bostondynamics.com. https://bostondynamics.com/about/history/
11. The Robot Report. "Boston Dynamics' Handle robot brings mobile manipulation to logistics." https://www.therobotreport.com/boston-dynamics-handle-robot-pallets/
12. designboom. "Boston Dynamics' 'Handle' robot is scarily good at parkour." 28 February 2017. https://www.designboom.com/technology/boston-dynamics-handle-robot-02-28-2017/
13. The Robot Report. "Ascento launches nimble Guard robot with wheel-leg design following a new $4.3M funding round." 2023. https://www.therobotreport.com/ascento-launches-nimble-guard-robot-following-a-new-4-3m-funding-round/
14. TechCrunch. "Ascento raises $4.3M for its two-wheeled security robot." 11 September 2023. https://techcrunch.com/2023/09/11/ascento/
15. MathWorks. "ETH Zurich Students Design and Build a Jumping Robot (Ascento)." https://ch.mathworks.com/company/mathworks-stories/eth-zurich-students-design-and-build-jumping-robot.html
16. The Robot Report. "Ascento Guard 2.0 is ready for security duty at large facilities." https://www.therobotreport.com/ascento-guard-2-0-ready-security-duty-large-facilities/
17. DEEP Robotics. "DEEP Robotics introduces LYNX M20, a new benchmark for industrial-grade wheel-legged robots." https://www.deeprobotics.us/news/deep-robotics-introduces-lynx-m20-a-new-benchmark-for-industrial-grade-wheel-legged-robots/
18. Generation Robots. "Unitree B2-W wheeled quadruped robot." https://www.generationrobots.com/en/404199-b2-wheeled-quadruped-robot.html
19. The Robot Report. "LimX Dynamics launches W1 wheeled quadruped." https://www.therobotreport.com/limx-dynamics-launches-w1-wheeled-quadruped/
20. New Atlas. "Wheeled legs let LimX Dynamics' W1 quadruped robot walk n' roll." https://newatlas.com/robotics/limx-dynamics-w1-wheeled-legs-quadruped-robot/
21. ROBOTS: Your Guide to the World of Robotics. "EVE (1X / Halodi Robotics)." https://robotsguide.com/robots/eve
22. NASA Jet Propulsion Laboratory. "RoboSimian." https://www.jpl.nasa.gov/robotics-at-jpl/robosimian/
23. Interesting Engineering. "Swiss-UK team rolls out robot dogs to transform doorstep deliveries." https://interestingengineering.com/innovation/autonomous-wheeled-legged-robots-tackle-last-mile-parcel-delivery
24. ETH Zurich Robotic Systems Lab. "ANYbotics spin-off." https://rsl.ethz.ch/partnership/spinoff/anybotics.html