Enabling Real-Time Shape-Sensing in Soft Robots via a Miniaturized, Single-Signal, Color-Tuned Soft Optical Sensor

Soft robotic systems require embedded sensing for closed-loop and autonomous control; however, existing shape-sensing approaches are limited by rigidity, bulky multisensor designs, and high computational cost. Here, we present Light-based Ultrathin Miniature Optical Sensor (LUMOS), a miniaturized, fully soft optical sensor with a 1.25 mm cross-section that utilizes thin film color tuning and a flexible mirror tip to deliver real-time 3D shape information from a single red-green-blue (RGB) response. The mirror enables full signal reflection and transmission from the base, eliminating the need for onboard electronics and allowing for miniaturization. Additionally, thin film color tuning adds minimal thickness and encodes sufficient data for complete 3D shape reconstruction. Integrated into a miniaturized soft continuum robot, LUMOS achieved a predicted tip accuracy of 0.47 mm using neural network mapping. In closed-loop operation, the robot demonstrated precise 3D targeting, maintained accuracy under external disturbances, and autonomously navigated in an in vitro lung phantom.