Robust detection problem is presented in this paper. The design of optimal decision rules in detection problems requires the knowledge of the conditional probability distributions of the observations given by each hypothesis. In many applications, however, the probability distributions are not specified completely. In these cases, the problem could be solved to find least favorable distributions. In the view of data fusion, the decentralized detection is better in information transmission and fault tolerance properties of the system. Finally,the data compression for multisensor system is discussed when ��N�� sensors work at different sampling rate.