Provided are a data processing method for multi-sensor fusion, a positioning apparatus and a VR device. The method includes: calculating a pose of a physical center by using data of at least one sensor set in a first positioning period; obtaining the predicted pose of the physical center according to an inertial navigation of the pose of the physical center; calculating a three-dimensional coordinate pose of each sensor set in a second positioning period; calculating, based on the three-dimensional coordinate pose, third inertial navigation data of each sensor set in the second positioning period; obtaining the converted poses of the physical center respectively through the third inertial navigation data and a pose relationship between each sensor and the physical center; and obtaining the predicted pose of each sensor set based on the converted poses of the physical center.