Holiday Robotics Friday

Scroll sideways for more →

Friday is a wheeled humanoid robot developed by Holiday Robotics, a South Korean robotics startup founded by serial entrepreneur Song Ki-young. Standing 176 cm (5 ft 9 in) tall and weighing 115 kg, Friday features 63 degrees of freedom across its body, including highly dexterous 20-DOF hands with integrated tactile and force-torque sensors. The robot uses a wheeled mobile base rather than bipedal legs, a deliberate design choice that prioritizes stability, battery life, and manipulation capability over the complexity of walking.

Friday is positioned as an industrial-grade humanoid robot targeting advanced manufacturing, logistics, and research applications. Holiday Robotics markets the robot as "the world's most advanced humanoid robot with 63 DoF and dexterous hands," and has set a target price of approximately 100 million KRW (roughly $70,000 USD), which the company says is equivalent to the annual cost of a manufacturing worker in South Korea. The robot is powered by a proprietary AI architecture called Vision-Language-Skill (VLS) and trained using a custom simulation platform called Holiday Sim.

Holiday Robotics

Company background

Holiday Robotics (Korean: 홀리데이 로보틱스) was founded in April 2024 in Seoul, South Korea. The company is headquartered at 70 Nonhyeon-ro 85-gil, Gangnam-gu, Seoul. Its stated mission is "Liberating people from physical labor with versatile robots." The company tagline, displayed on its GitHub page, reads: "We Value Humans. We Create Humanoids. We Make Holidays."

Holiday Robotics is a private robotics engineering firm dedicated to integrating advanced artificial intelligence with physical hardware for complex real-world tasks. The company focuses on five primary areas: autonomous and AI systems, consumer applications, edutainment and research, humanoid and social robotics, and industrial manufacturing ^[1].

Founder: Song Ki-young

The company was founded by Song Ki-young (also romanized as Song Kiyoung), a prominent figure in South Korea's AI and technology industry. Song studied at Seoul National University and co-founded SUALAB in 2013, a Seoul-based startup that developed deep learning-based machine vision software for industrial inspection applications. SUALAB's SuaKit software used neural networks to detect atypical and irregular defects that traditional rule-based machine vision systems could not reliably identify. The company's clients included major manufacturers such as Samsung and LG, who entrusted SUALAB with some of their most challenging industrial inspection tasks in the display, solar, PCB, film, and semiconductor industries ^[2].

In October 2019, Cognex Corporation, a leading American machine vision company, acquired SUALAB for $195 million. The acquisition was described as the largest foreign acquisition of a Korean tech startup at the time. Following the deal, Song joined Cognex as Senior Director of Engineering, where he helped lead what the company called the world's largest team of engineers specializing in deep learning for industrial machine vision ^[3][4]. Song's experience building AI-powered industrial inspection systems directly informs Holiday Robotics' focus on precision manipulation and factory-floor applications.

Funding

Holiday Robotics secured a seed funding round of 17.5 billion KRW (approximately $12.9 million USD) in August 2024, just four months after the company was founded. The round was led by several prominent South Korean institutional investors, including Atinum Investment, Stonebridge Ventures, Spring Camp, and ZERO1NE Ventures. Choi Dong-yeol, head of investment at Stonebridge Ventures, led the investment ^[5][6].

The estimated enterprise value of Holiday Robotics at the seed stage ranged from 53 million to 80 million euros, an unusually high valuation for a seed-stage robotics company, reflecting investor confidence in Song's track record and the growing enthusiasm for humanoid robotics globally ^[1].

With the seed capital, Holiday Robotics announced plans to recruit top-level talent in robotics and AI, accelerate development of the Friday humanoid robot, and establish a robotics research institute in the United States by the first half of 2025 ^[5].

Partnerships

Holiday Robotics has formed a strategic partnership with Endotlight, a generative 3D AI startup. The collaboration focuses on enhancing the efficiency of robot AI training by automatically generating 3D synthetic datasets based on real work scenarios. This approach is intended to improve robot perception and control precision in complex industrial environments while reducing the cost and time associated with collecting real-world training data ^[1].

Design philosophy

Wheels over legs

Friday's most distinctive design decision is its use of a differential drive wheeled base instead of bipedal legs. This choice runs counter to the prevailing trend in the humanoid robotics industry, where companies such as Boston Dynamics, Tesla, Figure AI, and Unitree Robotics have focused heavily on bipedal locomotion. Holiday Robotics argues that the "iPhone moment" for robotics depends on dexterity and return on investment, not on walking.

Song Ki-young has explained the rationale: "Chinese companies are doing really well at walking. But in industrial sites, most of the value comes from the hands. We focused on the precision of 'grasping, turning, and feeling' rather than walking." The wheeled base provides several practical advantages over legs for industrial deployment: greater stability on factory floors, longer battery life due to lower energy consumption, higher movement speed, and the ability to operate 24/7 with hot-swappable batteries ^[7].

Holiday Robotics has not abandoned bipedal locomotion entirely, however. The company views the wheeled base as the fastest route to commercial viability and has announced plans to test a bipedal version of Friday as part of its longer-term roadmap. This approach, also seen in robots such as Unitree's G1-D, reflects a growing trend in the industry toward prioritizing stability and manipulation capability in initial commercial products while developing legged variants in parallel ^[7].

Hands-first approach

Friday's design philosophy centers on the belief that manipulation capability, particularly hand dexterity, is the primary bottleneck in industrial automation. While many humanoid robot manufacturers allocate significant engineering resources to locomotion, Holiday Robotics has invested disproportionately in its hand and arm systems. The 20-DOF hands are the robot's most technically ambitious subsystem and represent the core of its value proposition for manufacturing customers who need robots capable of intricate assembly, gripping, and tool-use tasks ^[7].

Technical specifications

Physical dimensions

Friday stands 176 cm (5 ft 9 in) tall and weighs 115 kg with the mobile base. The total weight is distributed as follows: the mobile base accounts for 66 kg, providing a low center of gravity for stability; the upper body weighs approximately 15 kg; each arm weighs roughly 5 kg; and the waist section accounts for approximately 24 kg. The shoulder width is estimated at 60 cm, with a depth of approximately 50 cm ^[8].

Scroll sideways for more →

Degrees of freedom

Friday has 63 total degrees of freedom, one of the highest DOF counts among current humanoid robots. The DOF are distributed across the body as follows:

Scroll sideways for more →

The 7-DOF arms provide human-like reach and flexibility, enabling the robot to access workspaces from multiple angles. The 5-DOF torso allows bending, rotation, and tilting movements that extend the effective workspace of the arms and allow the robot to lean into tasks. Combined with the 20-DOF hands, this configuration gives Friday a level of upper-body dexterity that exceeds most competing humanoid platforms.

For comparison, the Tesla Optimus has approximately 28 degrees of freedom, the Boston Dynamics Atlas (electric version) has approximately 28, and the Unitree G1 ranges from 23 to 43 depending on configuration. Friday's 63 DOF total is significantly higher than all of these, with the difference concentrated in its hands.

Hand system

The hands are Friday's most technically distinctive feature. Each hand provides 20 degrees of freedom and weighs only 500 grams (approximately 1.1 lb), making them lightweight enough to minimize the load on the arm actuators while providing fine manipulation capability.

Key specifications of the hand system include:

Scroll sideways for more →

The hands feature integrated tactile and force-torque sensors that provide full-palm contact detection. Unlike many robotic hands that place tactile sensors only on fingertips, Friday's sensors cover both fingers and palms, enabling the robot to sense contact across the entire grasping surface. The sensors can detect forces as light as 0.05 N (approximately 5 grams) at a refresh rate of 1,900 Hz, providing high-resolution, real-time feedback for compliant grasping and delicate manipulation tasks ^[7][8].

Holiday Robotics uses a cost-effective magnetic sensor design on silicon substrates, manufactured at approximately 30,000 KRW (roughly $21 USD) per sensor unit. This low per-unit cost allows the company to deploy sensors densely across the entire hand surface rather than limiting them to a few critical contact points, a significant advantage for complex manipulation tasks ^[7].

The 20-DOF configuration enables the hands to perform intricate assembly operations, tool use, and object manipulation that require high-level precision. This DOF count places Friday's hands among the most dexterous in the humanoid robotics industry, comparable to the Shadow Dexterous Hand (20 DOF) and the Linkerbot Linker Hand L20 (20 DOF), both of which are standalone research platforms rather than integrated components of a complete humanoid system.

Manipulation capability

Friday's manipulation specifications are designed for industrial workloads:

Scroll sideways for more →

The 5 kg per-arm payload is suitable for handling components, tools, and products commonly encountered in electronics manufacturing and light assembly. The 20 kg proximal payload capacity (achieved when the load is close to the body) enables heavier lifting tasks such as moving boxes and stacking parts. The reported maximum lifting capacity of 68 kg is achieved through coordinated dual-arm lifting in optimal configurations ^[8].

Mobility

Friday uses a differential drive wheeled mobile base that enables rapid, stable movement across flat indoor surfaces.

Scroll sideways for more →

The wheeled base provides a maximum speed of 1.9 m/s, which is competitive with bipedal humanoid robots and significantly faster than many legged platforms in practice. For context, the Unitree H1 holds the speed record for humanoid robots at 3.3 m/s, but most bipedal humanoids operate at walking speeds well below 2 m/s during practical tasks. The 66 kg base provides a low center of gravity that keeps the robot stable during manipulation tasks, preventing tipping when the arms are extended with heavy loads ^[7][8].

Sensor suite

Friday is equipped with a comprehensive multi-modal sensor system for perception, navigation, and manipulation feedback:

Scroll sideways for more →

The combination of LiDAR and stereo cameras enables simultaneous localization and mapping (SLAM) for autonomous navigation, while the hand-mounted tactile sensors provide the fine-grained force feedback necessary for delicate manipulation tasks ^[8].

Computing and connectivity

Friday uses an NVIDIA Jetson Orin module as its primary onboard computing platform, providing the AI processing power required for real-time perception, decision-making, and motor control. The Jetson Orin delivers up to 275 TOPS (tera operations per second) of AI compute performance, enabling on-device machine learning inference for object detection, grasp planning, and navigation without relying on cloud connectivity for latency-sensitive operations ^[8].

Scroll sideways for more →

The system achieves a 0.2-second glass-to-action response time, meaning the delay from visual input to physical motor response is kept under 200 milliseconds. This low latency is critical for reactive manipulation tasks where the robot must adjust its grip in real time based on tactile and visual feedback ^[8].

Friday's control interface is accessible through a tablet or PC, allowing operators to switch between manual control, task programming, and autonomous mode activation.

Power system

Friday is designed for continuous industrial operation through a hot-swappable battery system:

Scroll sideways for more →

The hot-swappable battery design addresses one of the most significant practical limitations of humanoid robots in industrial settings: downtime for charging. By allowing batteries to be replaced without shutting down the robot, Friday can maintain continuous operation across multiple shifts. This feature gives it a significant advantage over bipedal humanoids, many of which struggle to achieve more than one to two hours of runtime on a single charge ^[7][8].

Operating environment

Scroll sideways for more →

Safety features

Friday includes multiple safety systems designed for collaborative operation alongside human workers:

• Force-limiting joints that prevent excessive contact forces
• Real-time collision detection and avoidance
• Emergency stop functionality
• Collaborative mode for safe human-robot interaction
• Compliant joint control to absorb unexpected contacts

These safety features are essential for deployment in manufacturing environments where the robot works alongside or near human workers, following collaborative robot safety standards ^[8].

Software and AI

Vision-Language-Skill (VLS) framework

Friday's AI system is built on a proprietary architecture called Vision-Language-Skill (VLS), which differs from the Vision-Language-Action (VLA) models used by many competing humanoid robot developers. Song Ki-young has explained the distinction: "VLA is a black box, so it's hard to predict and guarantee safety" ^[7].

In a standard VLA architecture, a single end-to-end neural network takes visual and language inputs and directly outputs raw motor commands. While powerful, this approach can produce unpredictable behaviors because the mapping from perception to action is opaque. Holiday Robotics' VLS framework takes a different approach: the AI system observes the environment through cameras and other sensors, interprets high-level instructions through natural language processing, and then selects from a library of pre-verified control "skills" (such as grasping, pushing, inserting, or turning) rather than generating raw motor commands from scratch ^[7].

This hybrid approach aims to combine the flexibility of generative AI (understanding open-ended instructions and adapting to novel situations) with the reliability and predictability of traditional control theory (executing well-tested motion primitives). The "whitebox" nature of the VLS system means that operators can inspect which skill the robot has selected and why, providing transparency that is critical for factory-floor deployment where safety and predictability are paramount.

Holiday Sim

Holiday Robotics has developed a proprietary simulation platform called Holiday Sim, designed to bridge the sim-to-real gap that has historically limited the transfer of robot behaviors from virtual training environments to physical hardware. Holiday Sim focuses specifically on accurately modeling soft-body contact dynamics, addressing limitations in widely used physics engines such as MuJoCo when simulating the complex interactions between robot hands and deformable or fragile objects ^[7].

Accurate simulation of soft-body dynamics is particularly important for training the tactile manipulation skills that are central to Friday's value proposition. If the simulation does not faithfully represent how objects deform under pressure, how surfaces slide against each other, or how forces propagate through contact, the policies trained in simulation will fail when transferred to the real robot.

FlashSAC: reinforcement learning research

Holiday Robotics maintains an active open-source presence on GitHub. The company's most prominent public repository is FlashSAC, described as "Fast and Stable Off-Policy Reinforcement Learning for High-Dimensional Robot Control." The project, released under an MIT license, has attracted 125 stars on GitHub as of April 2026. FlashSAC addresses the challenge of training reinforcement learning policies for robots with many degrees of freedom, directly relevant to controlling Friday's 63-DOF body ^[9].

The company also maintains several ROS 2 integration repositories, sensor and hardware drivers for inertial measurement units and laser scanning systems, and research tools for kinematics calibration and robot payload identification ^[9].

Product roadmap

Holiday Robotics' product naming convention reveals a broader expansion strategy beyond industrial automation:

Scroll sideways for more →

The naming reflects a week-of-work progression: Friday handles the hard industrial labor, Saturday moves into service-sector work, and Sunday comes home. Each product is expected to build on the core hand dexterity and AI platform developed for Friday while adapting the form factor and capabilities to its target environment ^[7].

Song Ki-young has articulated a survival-focused commercial philosophy: "The robot companies that survive 3-5 years from now will be the ones that commercialized." This urgency drives Holiday Robotics' focus on reaching production quickly with a practical, deployable product rather than pursuing technically impressive but commercially distant goals ^[7].

Production timeline

Holiday Robotics has announced plans for an initial manufacturing run of 100 units, with pilot programs already under discussion with automotive and electronics manufacturers. The company targets a price of 100 million KRW (approximately $70,000 USD) per unit, a figure calculated to be equivalent to the annual cost of a manufacturing worker in South Korea, enabling straightforward ROI calculations for potential industrial customers ^[7].

Applications

Manufacturing and assembly

Friday's primary target market is advanced manufacturing. The robot's 20-DOF hands with high-sensitivity tactile feedback are designed for tasks such as:

• Precision assembly of electronic components
• Tool use in manufacturing workflows
• Quality inspection through tactile sensing
• Parts handling and sorting
• Connector insertion and cable routing

The tactile sensors' ability to detect forces as light as 0.05 N enables the robot to handle delicate components without damaging them, while the 1,900 Hz sensor refresh rate provides fast enough feedback for dynamic manipulation tasks where grip force must be continuously adjusted.

Logistics

The wheeled mobile base and 20 kg proximal payload capacity make Friday suitable for logistics applications including:

• Package handling and sorting
• Warehouse pick-and-place operations
• Material transport within manufacturing facilities
• Inventory management and shelf stocking

The robot's 1.9 m/s maximum speed and 24/7 operational capability through hot-swappable batteries are particularly relevant for logistics environments that operate around the clock.

Research

Friday's high DOF count, comprehensive sensor suite, ROS 2 compatibility, and Python/C++ APIs make it a capable research platform for:

• Dexterous manipulation and grasping research
• Human-robot interaction studies
• Tactile perception and force control
• Mobile manipulation planning
• Reinforcement learning for high-DOF systems
• Sim-to-real transfer experiments

Human-robot interaction

Friday supports speech and gesture recognition capabilities, enabling interactive applications in demonstration settings, collaborative work environments, and service scenarios where the robot must communicate with human co-workers ^[8].

Competitive context

South Korean robotics landscape

Friday's development comes amid a significant surge in South Korean robotics activity. Major Korean conglomerates have entered the humanoid robotics space, and the government has launched initiatives to position the country as a global robotics leader:

Scroll sideways for more →

Holiday Robotics differentiates itself from these larger competitors through its startup agility, its specific focus on hand dexterity and industrial manipulation, and its founder's deep expertise in AI-powered industrial inspection ^[7].

Global humanoid robot comparison

Friday occupies a unique position in the global humanoid robot landscape due to its combination of a wheeled base and exceptionally high degrees of freedom:

Scroll sideways for more →

Friday's 63 total DOF and 20-DOF hands give it significantly more dexterity than any of the bipedal competitors listed. Its hot-swappable battery system and wheeled base enable continuous operation that bipedal robots currently cannot match. However, the lack of bipedal locomotion limits Friday's ability to navigate stairs, step over obstacles, and access spaces designed exclusively for walking humans.

Limitations

Despite its technical strengths, Friday has several notable limitations:

• Wheeled mobility restrictions: The wheeled base cannot navigate stairs, step over obstacles, or traverse highly uneven terrain. This limits deployment to environments with flat or gently sloped floors.
• Weight: At 115 kg, Friday is significantly heavier than most competing humanoid robots, which may complicate transport and limit deployment in some environments with weight restrictions.
• Early-stage product: As of early 2026, Friday remains in the development and pre-production phase. The robot has not yet been deployed at scale in industrial settings, and real-world performance may differ from specifications and demonstrations.
• Small company: Holiday Robotics had 1 to 10 employees at the time of its seed funding, significantly smaller than the engineering teams at Tesla, Boston Dynamics, or major Korean conglomerates entering the space.
• Limited track record: While Song Ki-young's SUALAB background demonstrates expertise in AI-powered industrial systems, the company has not yet demonstrated experience in volume manufacturing of complex robotic hardware.

See also

• Humanoid robot
• Humanoid robots
• Holiday Robotics
• Degrees of freedom
• Humanoid robot hands
• Sim-to-real transfer
• Reinforcement learning
• Machine vision
• ROS (Robot Operating System)
• NVIDIA Jetson
• Boston Dynamics
• Tesla Optimus
• Unitree G1
• Figure AI

References

1. "Holiday Robotics." Aparobot company profile. ↩
2. "Why Cognex Acquired SUALAB: Deep Learning." Cognex Blog. ↩
3. "Cognex Acquires SUALAB to Advance its Leadership in Deep Learning-Based Machine Vision." Business Wire, October 16, 2019. ↩
4. "Cognex acquires South Korean AI startup Sualab in $195 million deal." The Investor. ↩
5. "Holiday Robotics Secures KRW 17.5 Billion in Seed Funding for Humanoid Robot Development." WOWTALE, August 27, 2024. ↩
6. "Korea's AI, robot startups shine despite funding drought." KED Global, September 2024. ↩
7. "Holiday Robotics Unveils 'Friday': A Wheeled Humanoid That Prioritizes Hands Over Legs." Humanoids Daily. ↩
8. "Friday by Holiday Robotics: Price, Details, Review 2025." Origin of Bots. ↩
9. Holiday Robotics GitHub organization. ↩
10. "Friday - 63-DoF, dexterous humanoid." Humanoid Guide.
11. Holiday Robotics official website.
12. "South Korea: A prominent emerging market in AI robotics." Interact Analysis.