The trellis shaping is the method of selecting a minimum weight sequence from an equivalent class of possible transmitted sequences. It is perforemed by searching through the trellis diagram of a shaping convolutional code Cs. A shaping gain of the order of 1dB may be obtained with a simple 4-state shaping code. But when the rellis shaping is applied to multi-dimensional constellation, the shaping gain is decreased and the transmission delay is increased, which make it difficult to apply the trellis shaping to multi-dimensional constellation. In this paper, in order to solve these problems, the shell mapping algorithm is used for the mapper in the proposed trellis shaping. The simulatio result shows that the proposed trellis shaping is better than coventional trellis shaping to multi-dimensional constellation. In this paper, in order to solve these problems, the shell mapping on multi-dimensional costellation with respect to shaping gain and transmission delay. And the constellation expansion ratio (CER) and the capacity of transmitted injformation are identical to those of conventional trellis shaping on multi-dimensional constellation.