RobOmni Benchmark & Series A Funding: Daimon Robotics Builds Tactile Infrastructure for Embodied AI

At ICRA 2026, Daimon Robotics and Galbot jointly launched RobOmni, the world's first full-modal physical interACTion evaluation benchmark incorporating tactile sensing. Concurrently, the company announced the completion of a nine-figure Series A funding round, co-led by Inovance Industrial Investment and China Telecom.

By simultaneously releASIng an industry eValuation benchmark and SECuring strategic industrial capital, DAImon robotics is breaking free from the industry's entrenched "vision-only" perception cocoon. LeverAGIng its proprietary visuotactile Technology, the company is bridging the physical perception gap in Robotics, establishing a full-link "hardware-data-model" flywheel and steering the sector toward the next frontier of General-purpose dexterous manipulation.

Capital Bets on Tactile Sensing as a "Hard Necessity" for general embodied intelligence

Currently, the industry's computational power and resources are overwhelmingly tilted toward large vision models and VLA Frameworks. The deployment of Embodied Intelligence is hitting a critical disconnect between "seeing" and "doing": robots deliver impressive dEMOnstrations but fail to execute truly dexterous Operations in real-world settings.

This path dependency on pure vision traps robots in low-dimensional, rigid scenarios like material handling and sweeping. The industry remains stagnant before high-value processes requiring force and tactile physical interaction, such as precision assembly and flexible sorting.

The root cause lies in the constraint of tactile perception: robots can only "see shapes but not underStand physics." Confronted with slippage, occlusion, transparent objects, and flexible materials, they fail to achieve commercial deployment due to missing physical perception, trAPPing general-purpose Dexterous Manipulation in a demo dilemma.

Facing this perception-execution chasm, industrial automation giant Inovance technology, joined by China Telecom, co-led the investment in Shenzhen-based tactile sensing pioneer Daimon Robotics. As a core player deeply rooted in Industrial Automation and robotics infrastructure, Inovance's strategic move is no accident. It recognizes that multimodal systems lacking tactile closed-loop feedback are fundamentally incomplete. Only by filling the physical interaction perception puzzle can the industry unlock precision operation scenarios.

Daimon Robotics, incubated by the Hong Kong UniveRSIty of Science and Technology, is among the very few teams operating from a foundation of physical causality. Driven by massive visuotactile data, it is advancing physical world models for deployment in fine manipulation scenarios.

With this funding round, Daimon's capital landscape is taking clear shape. Previous rounds have attracted leading institutions including Kunlun Capital, JinDing Capital, Guozhong Capital, legend Capital, CMB International Capital, Oriental Fortune Capital, Bridge Capital, and China Merchants venture capital.

Coupled with the backing from these industrial investment giants, Daimon has rapidly constructed a dual-wheel ecosystem of financial investment and industrial capital, integrating core scenario resources across telecommunications, 3C electronics, and precision manufacturing. This is not a scatter-shot approach but a precise strategic encirclement of commercialization entry points for embodied tactile technology.

The composition of the investor syndicate reInforces an industry consensus: the large-scale deployment of industrial precision manufacturing and general-purpose robots absolutely cannot bypass the tactile physical closed-loop, which is the core key to bridging the Sim2Real gap. The rationale behind this intense investment lies in Daimon's globally pioneering monochromatic light visuotactile technology. Unlike traditional piezoelectric, capacitive, and multi-color optical tactile solutions, it encodes multidimensional tactile information directly into image data, seamlessly integrating with mainstream large vision model frameworks.

From the perspective of multimodal perception mechanisms, vision handles spatial localization, while tactile sensing anchors the causal closed-loop of physical interaction. When pure vision fails in long-tail scenarios involving transparent objects, flexible materials, and micro-slippage, tactile sensing compensates not just for blind spots but for the cognitive gap in physical causality. Without force and deformation feedback, robots can never truly comprehend the real world.

Robots lacking tactile sensing can only "see shapes, not grasp physical lAWS." Tactile sensing is the essential foundational modality for embodied intelligence to evolve from "seeing the world" to "operating in the world"—the core strategic rationale behind Daimon's unwavering focus on tactile sensing and physical interaction.

Dual Infrastructure Empowers the Commercialization of "Physical Intuition"

Global commercialization of tactile technology has long progressed slowly. The core bottleneck is not a single technical shortcoming but three interconnected obstacles: hardware manufacturing bARRiers, a scarcity of high-quality data, and the absence of standardized evaluation systems. This vicious cycle has been the primary barrier blocking the leap from laboratory validation to large-scale application for general-purpose dexterous manipulation.

First, hardware technology remains unconverged. Traditional piezoelectric and capacitive tactile sensors offer low Sampling point density. While technologically mature, their performance ceiling is limited, capturing only single-point pressure without the ability to capture deformation, slippage, or material texture. Meanwhile, traditional tri-color visuotactile sensors are bulky, power-hungry, and difficult to mass-produce.

Second, high-quality tactile data is scarce. Tactile data is generated through physical contact, making collection and annotation extremely costly. The industry has long lacked standardized, large-scale, full-modal datasets, starving model training of essential fuel.

Third, the industry lacks unified standards. There is no way to quantify how much tactile sensing improves robotic performance. Technical metrics remain fRAGmented across manufacturers, with hardware, data, and algorithms developing in isolation, resulting in blind R&D competition that fails to translate into real operational capability.

In short, overcoming the valley of death for tactile deployment requires winning three critical battles: breaking through in underlying sensor hardware, filling gaps in high-quality datasets, and establishing an evaluation benchmark. Following this logic, Daimon has anchored its strategy on proprietary monochromatic light visuotactile core technology, open-sourced the Daimon-Infinity dataset, and launched the RobOmni evaluation benchmark. The deployment of these two industry-grade infrastructures represents a structural deconstruction of the deadlock formed by the hardware-data-evaluation triad.

According to documentation, the Daimon-Infinity dataset, released in April 2026, is the world's largest full-modal physical world dataset incorporating tactile data. Positioned as a data engine, it effectively covers real industrial scenarios such as grasping, assembly, and flexible operations, filling the industry-wide scarcity of VTLA full-modal physical interaction data.

Simultaneously, Daimon is jointly building a data ecosystem with domestic and international universities, research institutions, and leading robotics enterprises. Open-sourcing portions of the dataset lowers industry R&D barriers, providing standardized, high-quality fuel for tactile model training and resolving the industry's tactile data shortage—answering the question of "where data comes from."

At ICRA in Vienna on June 3rd, Daimon and Galbot jointly launched RobOmni, the industry's first full-modal evaluation benchmark for physical interaction capabilities incorporating tactile sensing. Its launch directly addresses the long-standing pain point of embodied Intelligence lacking standardized metrics and unverifiable capability claims. With three core differentiated advantages, RobOmni is building irreplaceable industrial value defenses:

Evaluation dimension innovation: full-modal tactile sensing is incorporated into the benchmark. Breaking the industry's visual-only, success-rate-only evaluation paradigm, it brings tactile sensing into core assessment, focusing on contact-intensive fine manipulation. Beyond success rate, it adds physical metrics such as operational precision, robustness, and slip tolerance.

Quantifying tactile value, focusing on real-world deployment capability. Centered on task success rate and completion quality, it overlays multi-dimensional metrics including operational efficiency, robustness, and finesse. It uniquely supports tactile ablation studies, quantifying the performance gap with and without tactile sensing, using empirical data to prove the value of touch.

Closing the Sim2Real loop, it supports dual pathways of simulation training and real-world verification, achieving a bidirectional flow. It not only enables efficient model iteration within simulation environments but also achieves seamless transfer to real-world testing, establishing a "simulation optimization to real-world convergence" reinforcement pipeline, fundamentally resolving the industry's chronic issue of substantial gaps between physics engines and real tactile interaction.

As Daimon Robotics emphasizes its industrial vision: building the "embodied intelligence foundation iron triangle" through hardware, datasets, and evaluation benchmarks. But this is no mere technical assembly; it is a decisive tool to end the industry's era of standard-less, chaotic growth. Its implementation will forcefully push embodied intelligence beyond the "demo-first" performance phase and into a new industrial era driven by standards.

Predictably, through dataset open-sourcing and the dual infrastructure of the evaluation benchmark, an industry ecosystem flywheel encompassing "data accumulation, capability verification, and model iteration" is quietly taking shape. Daimon's strategic positioning in this move is precisely the "universal interface" for dexterous manipulation in the physical world.

From "Seeing" to "Feeling Accurately," Anchoring the Physical Universal Interface in the Second Half

In recent years, dexterous manipulation has raced down a misguided path of "hardware stacking and parameter inflation," yielding scattered results. Stripping away the surface, the fundamental bottleneck strangling this track is the "impossible triangle" that pure hardware approaches cannot transcend: high cost, poor reliability, and weak practical performance. These three constraints lock most end-user manufacturers out of mass production.

Daimon Robotics has carved an alternative path, using tactile sensing as the physical anchor to forge a closed-loop chain of synergistic evolution among "hardware, data, and models." Its core advantage lies in transforming execution end-effectors—such as dexterous hands—from mere parameter-obsessed mechanical structures into intelligent terminals genuinely governed by a "sensing, decision-making, execution" loop.

Here, hardware serves as the entry point for physical information acquisition, eschewing the obsession with single-parameter stacking in favor of rooting development in the fundamental pain point of tactile sensing for dexterous manipulation. Through self-developed visuotactile sensors, tactile grippers, and teleoperation collection devices, the company generates revenue through external market sales while continuously accumulating native visuotactile data across various deployment scenarios.

Real-scene data generated by physical robots, combined with synthetic data from RobOmni simulations, flows into the Daimon-Infinity dataset, continuously expanding VTLA full-modal data resources and becoming the core fuel for model iteration. Physical World Models are trained on this massive multimodal data, algorithm performance is continuously refined through standardized RobOmni evaluation, and the iteratively optimized models, in turn, guide the hardware iteration and upgrade of sensors, ultimately forming an endogenous triadic cycle.

On this foundation, Daimon Robotics has essentially constructed a growth flywheel driven by both data and commerce.

On the data flywheel side, performance evaluation data generated by partners using RobOmni will continuously flow back to enrich the underlying datasets. The leap in data scale and diversity directly enhances simulation fidelity and the authority of the evaluation system. This rising platform momentum will magnetically attract more ecosystem partners, forming a positive closed-loop of "data feedback, benchmark evolution, and ecosystem expansion," building a deep industrial moat.

On the commercial flywheel side, leveraging investor resources in industrial control, computing power, and industrial scenarios, Daimon's visuotactile hardware continues to be deployed in physical industries such as 3C precision processing, intelligent logistics, and component assembly. Commercial revenue continuously feeds back into cutting-edge R&D, while the customized demands generated by terminal deployment, in turn, drive the iterative upgrade of hardware, datasets, and evaluation benchmarks, continuously broadening the boundaries of commercialization.

Viewing the entire embodied intelligence industry lifecycle, the first half of competition focused intensely on large vision model development and robot volume production. The competitive foothold in the second half—general-purpose dexterous manipulation—lies in capturing the standardized universal interface for the physical world.

Daimon has established physical perception channels through self-developed tactile hardware, solidified the data foundation through open-sourced massive datasets, and established the evaluation benchmark through RobOmni, constructing an industry-wide universal tactile infrastructure in a triadic framework. The ultimate value of this architecture is essentially installing a "USB interface" onto the physical world.

Whether industrial robotic Arms or humanoid robots, breaking through laboratory limitations to achieve large-scale commercial deployment requires a standardized tactile interface for data acquisition, model training, and performance verification. When all dexterous manipulation operations are compelled to communicate based on this standard, Daimon will hold the foundational protocol for robotic physical interaction. As industry tactile data standards and evaluation benchmarks gradually unify, general embodied intelligence will formally enter a new phase of large-scale Physical AI Deployment centered on physical interaction.