Design and Control of a Sensorized End-Effector for Curvature-Based Haptic Recognition

This paper presents the design and control of a sensor-integrated three-finger end-effector for real-time curvature-based haptic recognition in robotic systems. The device, mounted on a commercial UR3 robotic arm, employs embedded optical sensors to detect contact timing during object exploration. An onboard microcontroller estimates local surface curvature from the time-differential response across the fingers, enabling closed-loop control with deterministic behavior and sub-10 ms latency at a 100 Hz sampling rate. A probability density–based classifier, trained on simulated data and validated on hardware, achieves 95% recognition accuracy on curvature-defined stimuli, with optical sensing proving more reliable and repeatable than mechanical alternatives. A virtual environment was implemented to replicate sensor dynamics and robot kinematics, supporting scalable data generation and seamless virtual-to-real transfer. The lightweight and reproducible framework demonstrates how time-differential haptic signals can be integrated into automation-oriented control systems, offering robust, low-latency surface perception for industrial tasks such as object classification, material sorting, and automated inspection.