 This paper proposes a novel approach to adapting simulated data to real-world conditions for powered lower-limb prosthesis users. It uses unsupervised domain adaptation to reduce the gap between simulated and real-world data, allowing for more accurate classification of five common terrains. This can help improve the performance of powered lower-limb prosthetics in real-world environments. This article was authored by Chu Hongqin, Kuan Zhenzhang, Yu Chuanleng, and others.