Daimon Robotics Debuts Tactile Infrastructure for Physical AI at IROS

The company’s presentation, themed "From Touch to Intelligence," centers on a three-layer architecture that integrates perception, data acquisition, and world modeling. The foundation rests on the DM-Tac series, vision-based sensors boasting over 110,000 sensing units capable of capturing high-frequency data across twelve tactile modalities. These sensors are supported by the Data-Nexus system, which streamlines the pipeline for generating interaction data, and the Daimon-TWM world model, which grounds robotic decision-making in physical feedback.

Demonstrating these capabilities publicly for the first time since its August launch, the Daimon-TWM model powered two autonomous tasks at the booth. In a bead-stringing demonstration, the system navigated small-scale precision and deformable materials without relying on pre-programmed motion sequences. A second task involved heat-transfer printing on canvas totes, requiring the robot to manage multi-step, long-horizon operations—such as film removal and hot stamping—by continuously adjusting its movements based on real-time tactile input. Founded in 2023 by researchers from The Hong Kong University of Science and Technology, the Shenzhen-based firm aims to shift robotics from simple observation to nuanced physical interaction.