Touch Ignites Physical Intelligence: Xense Robotics Showcases the Real Ability of Tactile Intelligence at ICRA 2026 | The AI Journal

Xense Robotics Technology (Shanghai) Co., Ltd. presented its full-stack tactile intelligence technology system at the IEEE International Conference on Robotics and Automation (ICRA) 2026 in Vienna, Austria. Held from June 2 to June 4, 2026, the event featured the company under the core proposition of Touch Ignites Physical Intelligence. The booth display covered tactile hardware perception, data acquisition, tactile world models, and VTLA models. A major highlight was the VTLA model-driven dual-arm long-horizon compliant manipulation demonstration. This showcase involved a realistic carton-forming task requiring flexible grasping, unfolding, alignment, and folding. Unlike traditional fixed-trajectory programming, the robot relied on real-time tactile and visual feedback to adapt to dynamic contact states without manual preset trajectories. This proved the ability of tactile intelligence to enable autonomous decision-making in unstructured scenarios involving deformable materials. Xense Robotics also debuted the TacCap-Gripper, a wearable two-finger tactile data acquisition device designed for physical intelligence. The device integrates high-precision visuo-tactile sensors, IMU units, and encoders to synchronize motion, vision, and touch data. Alongside this, the company exhibited a complete product matrix of multimodal tactile sensors adaptable for fingertips, grippers, and humanoid robots. These tools aim to fill gaps in the industry’s physical intelligence data loop. Founded in May 2024 by Daolin Ma, an Associate Professor at Shanghai Jiao Tong University, Xense Robotics focuses on enabling robots to interact with the physical world beyond visual understanding. The company plans to continue iterating sensing hardware and algorithms to support large-scale real-world deployment of embodied intelligence across industrial and humanoid applications. Xense Robotics positions tactile intelligence as a critical gateway for moving embodied systems from laboratory validation to industrial application. By addressing the limitations of vision-only perception regarding force and deformation, the showcased solutions offer a pathway for more reliable physical interaction in unstructured environments. While the technical demonstrations are promising, the scalability of these tactile data loops remains to be seen in broader commercial deployments. Future success will depend on whether the industry adopts these closed-loop ecosystems as standard infrastructure for next-generation robotics.