Abstract: We present an all-purpose bidirectional recurrent autoencoder aiming at the lack of the generality of the existing retargeting networks of motion data represented by joint position.The autoencoder can retarget the motion data from a source to any target character.The autoencoder is trained by motion data represented by joint position and the loss function defined by reconstruction error.After training,the hidden units and reconstructed motion of the corresponding source motion data are calculated by the autoencoder.Then,we impose the bone length constraints,foot trajectory constraints,the root joint position constraints and bone-to-bone angle constraints on the reconstructed motion,and the cost is projected back into the hidden-unit space and optimize the hidden units iteratively.The experimental results on CMU motion database show that the proposed autoencoder and four constraints can implement the retargeting of motion data represented by joint position,and the retargeting results have better effects on bone length error,bone-to-bone angle error,and end effector trajectory and smoothness.