Design and integration of a multi-axial tactile sensor for dexterous manipulation by humanoid robots for industrial applications

• Novel multi-axial tactile sensor design : We introduce a low-cost, easy-to-manufacture tactile sensor that provides both normal and shear force sensing, enabling humanoid robots to achieve human-like dexterity. • Hybrid multi-material architecture : By combining rigid PLA, soft PDMS, and flexible TPU, the sensor effectively decouples normal and shear forces into distinct channels, solving a major challenge in tactile sensing. • High-performance sensing : Experimental validation shows a sensitivity of 0.467 V/N for normal forces and above 1.1 V/N for shear forces, with a fine force resolution down to 3.79 gram-force and a dynamic bandwidth of 150 Hz. • Industrial integration & validation : The sensor was successfully embedded into the fingertips of the ARAtronica humanoid robot, enabling stable grasping, slip detection, and real-time haptic control during complex tasks such as CNC and lathe machine operation. • Towards intelligent haptics in robotics : This work bridges the gap between traditional robotic automation and intelligent, touch-enabled manipulation—bringing humanoid robots one step closer to becoming safe and capable collaborators in Industry 5.0 environments. For humanoid robots to become effective collaborators in industrial environments, the ability to perform dexterous manipulation through a human-like sense of touch is paramount. However, many existing tactile sensors are limited by high costs, complex fabrication, or an inability to measure multi-axis forces, particularly the shear forces crucial for stable grasping and slip detection. This paper presents the complete design, characterization, and integration of a novel, low-cost multi-axial tactile sensor based on a multi-material architecture combining rigid, flexible, and soft polymers. The sensor's design effectively decouples normal and shear forces into three distinct channels, providing a differential, 180-degree out-of-phase output for unambiguous directional shear sensing. Experimental characterization demonstrated high performance, including a normal force sensitivity of 0.467 V/N and shear sensitivities exceeding 1.1 V/N. This resulted in an excellent force resolution capable of detecting changes as small as 3.79 gram-force (0.037 N), with a dynamic operational bandwidth of up to 150 Hz. The sensor was successfully integrated into the fingertips of the ARAtronica humanoid robot, enabling it to perform complex industrial tasks, including the haptic-guided operation of a CNC machine and a lathe, which would be unreliable without tactile feedback. The results validate our sensor as a practical and effective solution for enabling haptic dexterity, bringing humanoid robots a crucial step closer to becoming safe and capable collaborators in the factories of the future.