Abstract: A new information fusion method,named Dezert-Smarandache theory (DSmT),which is extended from Bayesian theory and Dempster-Shafer theory (DST),is introduced to solve the problem of mobile robot map building in an unknown dynamic environment.The grid map method is adopted,and according to the characteristics of sonar sensors,a sonar sensor mathematical model is constructed based on DSmT.With the application of hybrid DSm model,i.e.,an evolving model of DSmT,a group of general basic belief assignment functions (gbbaf) are constructed to deal with the uncertain, imprecise and even highly conflicting information obtained with sonar sensors in the dynamic environment.At last,mobile robot Pioneer 2-Dxe is used to carry out experiments of map building with hybrid DSm model and DST,and the correla- tive 2D general basic belief assignment (gbba) map is constructed.The ichnography created with the hybrid DSm model is compared with the map built with DST,and the result verifies the validity of the hybrid DSm model in dynamic unknown environment,supplying a powerful theoretic evidence to process highly conflicting dynamic information.