Uncertain is inherent in moving object trajectories due to measurement errors or time-discretized sampling. Unfortunately, most previous research on trajectory pattern mining did not consider the uncertainty of trajectory data. This paper focuses on the uncertain group pattern mining, which is to find the moving objects that travel together. A novel concept, uncertain group pattern, is proposed, and then a two-step approach is introduced to deal with it. In the first step, the uncertain objects' similarities are computed according to their expected distances at each timestamp, and then the objects are clustered according to their spatial proximity. In the second step, a new algorithm to efficiently mining the uncertain group patterns is designed which captures the moving objects that move within the same clusters for certain timestamps that are possibly nonconsecutive. However the search space of group pattern is huge. In order to improve the mining efficiency, some pruning strategies are proposed to greatly reduce the search space. Finally, the effectiveness of the proposed concepts and the efficiency of the approaches are validated by extensive experiments based on both real and synthetic trajectory datasets.