Design and Optimization of a Soft Magnetic Tactile Sensor

This paper introduces SoftMag, an affordable and efficient soft magnetic sensor capable of detecting contact and normal force. The sensor incorporates a three-axis Hall sensor and four counter-symmetric deployed NdFeB magnets embedded within a deformable porous body. An analytical model for computing the flux density is established as the foundation for design and optimization, which are validated via a multi-physics simulation framework. To calibrate the sensor, we trained a K-nearest-neighbor classifier for position and a feed-forward neural network for force measurement. Experimental evaluation demonstrates a global full-scale output of 7.12 N with a static error of 9.52%, and a global accuracy of 100% for position estimation with a resolution of 10 mm, suggesting that SoftMag has a promising potential for providing tactile feedback in soft grippers.