HKUST Research Team Develops EmArm Robotic Arm With Whole-Arm Tactile Sensing

TechXplore reported on October 17, 2026, that a research team at the Hong Kong University of Science and Technology (HKUST) had developed the Embodied Sensorimotor Robotic Arm, or EmArm, which can detect touch across both the upper arm and forearm. By integrating soft optical tactile skin directly into the robot’s skeleton and combining it with a real-time decision-making algorithm, the system can reportedly sense contact and adaptively manipulate objects.

A design that extends tactile sensing across the whole arm could enable robots to handle objects and avoid contact without relying solely on their end effectors. According to HKUST’s report, a distinguishing feature is the ability to respond in real time while differentiating types of contact, suggesting a result aimed at conditions closer to real-world manipulation than laboratory research alone.

EmArm’s significance lies not simply in demonstrating a robot that can “feel,” but in extending tactile sensing beyond the end effector to the whole arm, including the upper arm and forearm, and using that information for real-time decisions. Research on robotic arms often centers on grasping accuracy and contact detection at the end effector, but this report highlights the integration of soft optical tactile skin into the skeleton and the system’s ability to identify types of contact. In other words, the research integrates input (contact), recognition (tactile analysis) and control (avoidance and manipulation), in an effort to bring a conventional mechanical arm closer to a body that responds to contact. This development addresses a more fundamental challenge — how to capture real-world contact data in detail — rather than the broader trend of implementing robots equipped with generative AI. Expanding tactile sensing affects not only on-board recognition accuracy but also safety and dexterity during contact, suggesting that the research centers on the integrated design of sensor placement and mechanical structure, rather than the algorithm alone. Based on publicly available information, no pricing or mass-production plans have been confirmed, and it is reasonable to regard the work at this stage as a research result. As this newspaper has no related coverage, it will not be linked to previous reporting. See the original source for details. The points to watch are: (1) the range of contact pressure and area the system can detect; (2) the conditions under which tactile recognition remains stable; and (3) whether assessments of reproducibility and safety for industrial applications are released.