The fingertip force sensor is the key for the complex task of the dexterous underwater
hand, in order to safely grasp an unknown object using the dexterous underwater hand and
accurately perceive its position in the fingers, a sensor should be developed, which can detect the
force and position simultaneously. Furthermore, this sensor should be used underwater. It is
difficult to employ the accustomed calibration method for the characteristic of the fingertip force
sensor, and the accustomed method is not able to assure the precision. A calibration method based
on RBF (Radial-Basis Function) neural network is introduced. Furthermore, the calibration system
and program are also designed. The calibration experiment of the sensor is carried out. The results
show the nonlinear calibration method based on RBF neural network assure the precision of the
sensor, which meets the demand of research on the underwater dexterous hand.