Commercial model
Quote-based sales
Research prototype; no commercial pricing or sales availability has been announced.. That usually means the final commercial package depends on deployment scope, services, or negotiated terms.
Robot dossier
Ace
Release
Apr 22, 2026
Price
Price TBA
Connectivity
0
Status
Prototype
Speed
Ball returns up to 19.6 m/s; mobile locomotion not applicable
Research
Prototype
Ace
Ace is Sony AI's autonomous table-tennis research robot for studying physical AI in fast, interactive tasks. The system combines event-based vision, high-speed cameras, and reinforcement-learning control to track ball position and spin with millisecond timing, then return shots through an eight-degree-of-freedom racket platform. Sony AI says Ace followed International Table Tennis Federation rules and scored wins against elite players, while the Nature paper describes it as a real-world autonomous system competitive with elite human table-tennis players. It is a research prototype rather than a commercial sports or home robot, but it is notable for pushing real-time perception and agile robot control toward professional-speed human interaction.
Listed price
Price TBA
Research prototype; no commercial pricing or sales availability has been announced.
Release window
Apr 22, 2026
Current status
Prototype
Sony AI
Last verified
May 10, 2026
Share this robot
Open a plain share composer on X or Bluesky for this robot profile.
Technical overview
Core specifications and system stack
A fast read on the mechanical profile, sensing package, and platform integrations behind Ace.
Technical Specifications
Height
Not officially disclosed
Weight
Not officially disclosed
Battery Life
Not officially disclosed
Charging Time
Not officially disclosed
Max Speed
Ball returns up to 19.6 m/s; mobile locomotion not applicable
Operational profile
How this robot is configured
Capabilities
10
Connectivity
0
Key capabilities
Autonomous Table-Tennis Match PlayProfessional and Elite Player CompetitionHigh-Speed Ball and Spin TrackingEvent-Based Vision PerceptionDeep Reinforcement Learning ControlReal-Time Shot SelectionOne-Armed Serve MotionTopspin and Slice Returns
Explore further
Benchmark set
Compare with similar robots
Shortcuts to the closest alternatives in the current ui44 set.
Research
iCub
Italian Institute of Technology
€250.000
Active
Release is 17y olderStatus differs: Prototype vs Active
Compare side by side
Research
DRC-HUBO+
KAIST
Price TBA
Prototype
Release is 11y olderTop speed is 18.8 m/s slower
Compare side by side
Research
KANGAROO
PAL Robotics
Price TBA
Active
Release is 1y olderStatus differs: Prototype vs Active
Compare side by side
Research
Roadrunner
Robotics & AI Institute
Price TBA
Prototype
Compare side by side
About the Ace
3Sensors10Capabilities
The Ace is a Research robot built by Sony AI. Ace is Sony AI's autonomous table-tennis research robot for studying physical AI in fast, interactive tasks. The system combines event-based vision, high-speed cameras, and reinforcement-learning control to track ball position and spin with millisecond timing, then return shots through an eight-degree-of-freedom racket platform. Sony AI says Ace followed International Table Tennis Federation rules and scored wins against elite players, while the Nature paper describes it as a real-world autonomous system competitive with elite human table-tennis players. It is a research prototype rather than a commercial sports or home robot, but it is notable for pushing real-time perception and agile robot control toward professional-speed human interaction.
Pricing has not been publicly disclosed — typical for robots still in development. See all Sony AI robots on the Sony AI page.
Spec Breakdown
Detailed specifications for the Ace
Height
Not officially disclosedAt Not officially disclosed, the Ace is sized for its intended operating environment and use cases.
Weight
Not officially disclosedWeighing Not officially disclosed, the Ace balances structural integrity with portability and maneuverability.
Battery Life
Not officially disclosedWith a battery life of Not officially disclosed, the Ace can operate for sustained periods before requiring a recharge. Battery life is measured under typical operating conditions and may vary based on workload intensity and environmental factors.
Charging Time
Not officially disclosedA charging time of Not officially disclosed means the ratio of operation to downtime is an important consideration for applications requiring near-continuous availability. Some deployments use multiple robots in rotation to maintain uninterrupted service.
Maximum Speed
Ball returns up to 19.6 m/s; mobile locomotion not applicableA top speed of Ball returns up to 19.6 m/s; mobile locomotion not applicable is calibrated for the robot's primary operating environment and safety requirements.
The Ace uses Deep reinforcement-learning table-tennis control trained in simulation, with low-latency event-based perception, 31.25 Hz policy updates, and 1 kHz trajectory execution on the robot hardware as its intelligence backbone. This AI platform powers the robot's decision-making, perception processing, and autonomous behavior. The sophistication of the AI stack directly impacts how well the robot handles unexpected situations and adapts to new environments.
Ace Sensor Suite
The Ace integrates 3 sensor types, forming the perceptual foundation that enables autonomous operation.
9 Sony Pregius IMX273 active-pixel cameras operating at 200 Hz
Details →
3 Sony IMX636 event-based vision sensors in gaze-control systems
Details →
Pan/tilt mirrors and telephoto tunable lenses for ball spin tracking
Details →
This sensor configuration enables the Ace to perceive its environment and operate autonomously in its intended use cases. Multiple sensor modalities provide redundancy and more robust perception than any single sensor type alone.
Explore sensor technologies: components glossary · full components directory
Ace Use Cases & Applications
Research robots serve as platforms for advancing robotics science and engineering. They enable researchers to test theories about locomotion, manipulation, perception, and human-robot interaction in controlled and real-world environments.
Capabilities That Enable Real-World Use
The Ace offers 10 distinct capabilities, each contributing to the robot's practical utility.
Autonomous Table-Tennis Match Play
Professional and Elite Player Competition
High-Speed Ball and Spin Tracking
Event-Based Vision Perception
Deep Reinforcement Learning Control
Real-Time Shot Selection
One-Armed Serve Motion
Topspin and Slice Returns
Collision-Checked Trajectory Planning
Sim-to-Real Robotics Research
These capabilities work together with the robot's 3 onboard sensor types and Deep reinforcement-learning table-tennis control trained in simulation, with low-latency event-based perception, 31.25 Hz policy updates, and 1 kHz trajectory execution on the robot hardware AI platform to deliver practical, real-world performance.
Ace Capabilities
10
Capabilities
3
Sensor Types
AI
Deep reinforcement-learning …
Autonomous Table-Tennis Match Play
Professional and Elite Player Competition
High-Speed Ball and Spin Tracking
Event-Based Vision Perception
Deep Reinforcement Learning Control
Real-Time Shot Selection
One-Armed Serve Motion
Topspin and Slice Returns
Collision-Checked Trajectory Planning
Sim-to-Real Robotics Research
Ace Technology Stack Overview
The Ace by Sony AI integrates 4 distinct technology components across sensing, connectivity, intelligence, and interaction layers. The physical platform features a height of Not officially disclosed, a weight of Not officially disclosed, a top speed of Ball returns up to 19.6 m/s; mobile locomotion not applicable, providing the foundation on which this technology stack operates.
Perception — 3 Sensor Types
The perception layer is built on 9 Sony Pregius IMX273 active-pixel cameras operating at 200 Hz, 3 Sony IMX636 event-based vision sensors in gaze-control systems, Pan/tilt mirrors and telephoto tunable lenses for ball spin tracking. These work in concert to give the robot a detailed understanding of its operating environment. This multi-sensor approach provides redundancy and enables the robot to function reliably even when individual sensors encounter challenging conditions such as low light, reflective surfaces, or cluttered spaces.
Intelligence — Deep reinforcement-learning table-tennis control trained in simulation, with low-latency event-based perception, 31.25 Hz policy updates, and 1 kHz trajectory execution on the robot hardware
Deep reinforcement-learning table-tennis control trained in simulation, with low-latency event-based perception, 31.25 Hz policy updates, and 1 kHz trajectory execution on the robot hardware serves as the computational brain, processing sensor data, making navigation decisions, and orchestrating the robot's autonomous behaviors. The quality of this AI platform directly influences how well the robot handles novel situations, adapts to changes in its environment, and improves its performance over time through learning.
Who Should Consider the Ace?
Target Audience
Research robots are acquired by universities, government labs, and corporate R&D departments. They serve as experimental platforms for developing new algorithms, testing locomotion strategies, and advancing the field of robotics. Some are also used for educational purposes.
Key Considerations
Open-source software compatibility (ROS/ROS 2), sensor modularity, programmability, available SDK/API quality, community support, and published research papers using the platform are key factors. Documentation quality and the ability to modify both hardware and software are essential for research use.
Pricing
Ace does not currently have publicly listed pricing. As the robot is still in development, pricing will likely be announced closer to market availability.
Availability
PrototypeThe Ace is currently in the prototype stage. It is not yet available for purchase, and specifications may change before the final product is released.
Ace: Strengths & Trade-offs
Engineering compromises and where this research robot excels
What the Ace does well
Broad capability set
With 10 distinct capabilities, the Ace is designed as a versatile platform rather than a single-task device. This breadth means the robot can handle varied scenarios and workflows, reducing the need for multiple specialized robots and increasing its utility across different situations.
Strong mobility performance
A top speed of Ball returns up to 19.6 m/s; mobile locomotion not applicable provides the Ace with the agility to cover ground efficiently. This is particularly valuable for applications that require rapid response, large-area coverage, or keeping pace with human movement in shared environments.
What to consider carefully
Undisclosed pricing
Sony AI has not published a public price for the Ace. While common for enterprise-class robotics, the absence of transparent pricing can complicate budgeting and comparison shopping. Prospective buyers will need to engage directly with the manufacturer for quotes, which may vary by configuration and volume.
Currently in prototype
The Ace is not yet available as a finished, shipping product. Specifications may change before commercial release, and timelines for availability are subject to revision. Early adopters should account for this uncertainty in their planning.
Limited ecosystem integration info
No specific smart home or ecosystem compatibility is listed for the Ace. This does not necessarily mean the robot lacks integration options — the information may not yet be published — but buyers who rely on specific platforms (Apple HomeKit, Google Home, Amazon Alexa, etc.) should verify compatibility before purchasing.
Note: This strengths and trade-offs assessment is based on the Ace's documented specifications as tracked in the ui44 database. Real-world performance depends on deployment conditions, firmware maturity, and environmental factors. For the most current information, check the Sony AI manufacturer page or visit the official product page. Use the comparison tool to evaluate these trade-offs against competing robots in the same category.
How Research Robot Technology Works
Understanding the engineering behind this category
Research robots serve a fundamentally different purpose than commercial or consumer models. They are platforms for discovery — enabling scientists and engineers to test theories, develop algorithms, and push the boundaries of what robots can do. The technology in research robots prioritizes openness, flexibility, and access to raw data over consumer-friendly packaging or commercial reliability. Understanding this distinction is important for anyone considering a research robot platform.
Navigation & Mobility
Research robots typically expose their navigation systems at a much lower level than commercial products. Researchers can access raw sensor data, modify SLAM algorithms, implement custom path planners, and test novel navigation approaches. ROS (Robot Operating System) and ROS 2 compatibility is standard, providing a common framework for sharing navigation modules across the research community. This openness enables rapid iteration — a researcher can swap between different SLAM implementations, test new obstacle avoidance strategies, or develop entirely novel navigation paradigms without being locked into a vendor's proprietary stack.
The Role of AI
Research robots serve as physical testbeds for AI algorithms that may eventually appear in commercial products years later. Reinforcement learning, imitation learning, few-shot task learning, and human-robot interaction studies all require robot platforms that can execute AI-generated commands in the physical world. The gap between simulation (where training is cheap and fast) and reality (where physics is unforgiving) makes physical robot platforms essential for validating AI approaches. Research robots must support rapid deployment of new AI models without extensive integration work.
Sensor Fusion & Perception
Research platforms prioritize sensor modularity and data access. Standard mounting interfaces allow researchers to attach custom sensors alongside built-in ones. Raw sensor data streams (not just processed results) are accessible for developing novel perception algorithms. Precise time-stamping and synchronization across sensor streams enable accurate multi-modal fusion research. Many research robots include more sensors than strictly necessary for any single application, providing researchers with rich datasets for developing and testing new algorithms.
Power & Battery Management
Research robots balance operational runtime with practical lab use. Sessions of one to four hours are typical, with quick charging between experiments. Some research setups use tethered power for long-running experiments where battery limitations would interrupt data collection. Power monitoring and logging capabilities help researchers understand the energy costs of different behaviors and algorithms — important for developing efficient approaches that will eventually run on battery-constrained commercial systems.
Safety by Design
Research environments present unique safety challenges because robots are constantly being programmed with untested behaviors. Hardware safety limits (joint speed caps, force limits, emergency stops) must be robust regardless of software commands. Safety-rated monitored stop and speed monitoring ensure the robot cannot exceed safe operating parameters even when running experimental code. Collaborative operation standards apply when researchers work alongside the robot during experiments. Many labs implement layered safety with physical barriers for high-speed testing and open-area operation restricted to validated, lower-risk behaviors.
What's Next for Research Robots
Research robot platforms are becoming more accessible and capable. Cloud robotics enables remote experiment execution and shared datasets. Digital twins and high-fidelity simulators reduce the need for physical hardware time while improving sim-to-real transfer. Standardized benchmarks and open datasets enable fair comparison of results across labs. The democratization of robotics research — through lower-cost platforms, open-source software, and cloud infrastructure — is expanding who can contribute to advancing the field.
The Ace by Sony AI incorporates many of these technology pillars. For a detailed look at the specific sensors and components used in the Ace, see the sensor analysis and connectivity sections above, or browse the complete components glossary for explanations of every technology used across the robotics industry.
Ace in the Research Market
How this robot compares in the research landscape
Sony AI has not publicly disclosed pricing for the Ace, which is typical for enterprise-focused robotics platforms that offer customized solutions and direct-sales relationships.
The Ace's 3 sensor types provide solid perceptual coverage for its intended use cases. This mid-range sensor suite balances cost with capability, covering the essential modalities needed for research applications.
As a robot still in prototype, the Ace represents Sony AI's vision for where research robotics is heading. Specifications may evolve before commercial release, and early performance demonstrations should be evaluated with this context in mind.
Head-to-Head Comparisons
Side-by-side specs, capability overlap analysis, and key differentiators.
For the full picture of Sony AI's portfolio and market strategy, visit the Sony AI manufacturer page.
Deployment Readiness and Procurement Signals for Ace
What the public profile tells you, and what still needs direct vendor confirmation
From a buying and rollout perspective, the Ace should be read as a research platform aimed at labs and development teams validating robotics workflows. ui44 currently tracks 10 capability signals, 3 sensor inputs, and a last verification date of 2026-05-10. That mix gives buyers a useful first-pass picture, but it is still only the public layer of due diligence, especially when procurement, uptime, and support commitments are decided directly with Sony AI.
Integration posture
Integration details thin
The page does not list any connectivity standards, so procurement teams should verify network requirements, remote management options, and how the robot fits into existing software or facility infrastructure.
Spec disclosure
1/7 core specs public
ui44 currently has 1 of 7 core physical and operating specs filled in for this model, leaving 6 gaps that matter for deployment planning. Missing runtime, charge, speed, or payload details can materially change staffing and site-readiness assumptions.
The current profile is useful for scouting, but it still leaves meaningful operational unknowns. If this robot is heading toward a pilot or purchase discussion, the next step should be a structured vendor Q&A that fills the remaining runtime, charging, payload, safety, or integration blanks before anyone builds ROI assumptions around it.
If you want a faster apples-to-apples read, compare the Ace against nearby alternatives in ui44's compare view, then cross-check the underlying AI, sensor, and subsystem terms in the components glossary. For manufacturer-level context, the Sony AI profile helps anchor this robot inside the wider product lineup.
Before you sign off on a pilot, confirm these points
- Ask for real shift runtime under the intended workload, not just standby endurance.
- Confirm how the charging workflow works in practice, including charger count, swap options, and expected downtime.
- Clarify usable payload or tool-load limits before planning material handling or mounted accessories.
- Request concrete API, integration, or workflow examples instead of assuming the robot will drop into an existing stack.
Owning the Ace: Setup, Maintenance & Tips
Practical guide from day one through years of ownership
Initial Setup
Research robot setup combines hardware assembly with software environment configuration. Unpack and assemble the platform following the manufacturer's documentation. Install the development framework — typically ROS or ROS 2 — and verify sensor connectivity. Calibrate all sensors using the manufacturer's tools and procedures. Set up the simulation environment (Gazebo, Isaac Sim, or equivalent) alongside the physical platform for parallel development. Establish version control for your experiment code and configuration. Document the initial calibration values and system state as your baseline for future reference. Plan network and computing infrastructure to handle the data rates your sensors will generate.
Ongoing Maintenance
Research robots need maintenance that preserves the precision required for valid experimental results. Regularly verify sensor calibration — drift in camera intrinsics or IMU biases can invalidate experiment data. Maintain clean workspace conditions to protect optical sensors. Document any hardware modifications or maintenance performed, as these can affect experimental reproducibility. Update software dependencies carefully, documenting versions used for each experiment. Joint and actuator wear in research robots that perform repetitive tasks should be monitored and factored into experimental design.
Software Updates & Long-Term Support
Research robot software updates require careful management to maintain experiment reproducibility. Document the exact software versions used for each experiment. Test updates in a separate environment before applying to your experiment platform. Contribute bug fixes and improvements back to the community when using open-source frameworks. Be aware that ROS and other framework updates may require code changes in your custom packages — budget time for integration testing after major framework updates.
Maximizing Longevity
Research robots often have longer productive lives than commercial products because they can be upgraded and repurposed. Extend your investment by maintaining clean mechanical and electrical systems, documenting all modifications for future lab members, and keeping spare parts for common wear items. When specific components become obsolete, community forums and lab networks can be valuable sources for replacements. Consider the platform's modularity when planning future research directions — a platform that can accept new sensors and actuators adapts to evolving research questions.
For Sony AI-specific support resources and documentation, visit the Sony AI page on ui44 or check the manufacturer's official website at Sony AI's product page.
Frequently Asked Questions
What is the Ace?
The Ace is a Research robot made by Sony AI. Ace is Sony AI's autonomous table-tennis research robot for studying physical AI in fast, interactive tasks. The system combines event-based vision, high-speed cameras, and reinforcement-learning control to track ball position and spin with millisecond timing, then return shots through an eight-degree-of-freedom racket platform. Sony AI says Ace followed International Table Tennis Federation rules and scored wins against elite players, while the Nature paper describes it as a real-world autonomous system competitive with elite human table-tennis players. It is a research prototype rather than a commercial sports or home robot, but it is notable for pushing real-time perception and agile robot control toward professional-speed human interaction. It features 3 sensor types, 0 connectivity protocols, and 10 distinct capabilities.
How much does the Ace cost?
Sony AI has not disclosed public pricing for the Ace. Pricing is typically announced closer to market release. Research prototype; no commercial pricing or sales availability has been announced.
Is the Ace available to buy?
The Ace currently has a status of Prototype. Check with Sony AI for the latest availability.
What sensors does the Ace have?
The Ace is equipped with 3 sensor types: 9 Sony Pregius IMX273 active-pixel cameras operating at 200 Hz, 3 Sony IMX636 event-based vision sensors in gaze-control systems, Pan/tilt mirrors and telephoto tunable lenses for ball spin tracking. These sensors work together through sensor fusion to provide comprehensive environmental awareness for autonomous operation. See the sensor analysis section for details.
How long does the Ace battery last?
The Ace has a rated battery life of Not officially disclosed and charges in Not officially disclosed. Actual battery performance may vary based on usage intensity, ambient temperature, and specific tasks being performed. Heavy workloads like continuous navigation and sensor processing will consume battery faster than idle or standby modes.
What AI does the Ace use?
The Ace is powered by Deep reinforcement-learning table-tennis control trained in simulation, with low-latency event-based perception, 31.25 Hz policy updates, and 1 kHz trajectory execution on the robot hardware. This AI platform handles the robot's perception processing, decision-making, and autonomous behavior. The sophistication of the AI directly impacts how well the robot handles unexpected situations, learns from its environment, and improves over time.
How does the Ace compare to the iCub?
The Ace and iCub are both research robots, but they differ in key specifications, pricing, and manufacturer approach. Use the side-by-side comparison tool to see detailed differences in specs, sensors, and capabilities. You can also browse other similar robots below.
How current is the Ace data on ui44?
The Ace specifications on ui44 were last verified on 2026-05-10. All data is sourced from official Sony AI documentation, spec sheets, and press releases. If you notice any outdated information, please let us know.
Data Integrity
All Ace data on ui44 is verified against official Sony AI sources, including spec sheets, product pages, and press releases. Last verified: 2026-05-10. Official source: Sony AI product page. If you find outdated or incorrect information, please let us know — accuracy is our top priority.
Explore More on ui44
Manufacturer
Category
Explore more research robots
See how the Ace stacks up — compare specs, browse the research category, or search the full database.