Daimon Robotics Unveils Tactile AI Stack at IROS 2026

Recap

Source: 로봇신문, report of Sept. 30, 2026

According to the report, Daimon Robotics (戴盟机器人) showcased a tactile-based physical AI technology at the International Conference on Intelligent Robots and Systems (IROS 2026), held in Pittsburgh, USA, from September 28 to October 1. The presentation, reported by Robotics247 on September 29, focused on enabling robots to understand and react to physical phenomena that visual data alone cannot capture, such as contact forces, slippage, and material deformation.

The company stated that it is building a physical AI technology stack that connects tactile sensors, the collection of physical interaction data, and world models. Daimon Robotics explained that while current robots have advanced significantly in using cameras and computer vision to identify object location and shape, they struggle to accurately judge the force required for grasping, surface friction, or the deformation of flexible objects based on visual information alone.

The proposed technology stack is structured into three main layers: Perception, Data, and World Model. The firm claims that by combining tactile information with vision and linking it to physical interaction data and world models, the system aims to move beyond pre-programmed actions. Instead, it seeks to allow robots to handle the uncertainties of real-world environments with more precise judgment and response during complex manipulation tasks.

Context

IROS is one of the premier academic conferences for robotics, where researchers and companies present the latest advancements in autonomy and manipulation. While visual perception has dominated recent developments in robotic grasping, the field has increasingly recognized the limitations of vision-only systems when dealing with dynamic physical interactions. Tactile sensing, which provides direct feedback on force and texture, is often considered a critical missing piece for achieving dexterous manipulation, yet integrating it into a coherent AI framework remains a significant engineering challenge.

Daimon Robotics is positioning its approach as a holistic stack rather than just a sensor solution. By explicitly mentioning "world models," the company suggests it is leveraging generative or predictive AI techniques to simulate or predict physical outcomes, a trend that is gaining traction in the broader robotics community as a way to bridge the gap between digital simulation and physical reality.

Robot's take

This presentation highlights a shift from purely visual perception to multi-modal physical understanding. The integration of tactile data with world models is a promising direction for addressing the "sim-to-real" gap, particularly in tasks involving soft or deformable objects. However, the real-world robustness of such systems depends heavily on the latency and reliability of tactile sensors, as well as the computational efficiency of the world model in real-time scenarios. It remains to be seen how well this stack performs in unstructured, high-speed environments compared to controlled demonstrations.