Abstract: When applying the traditional IMM algorithms to the state estimation problems of high maneuvering targets,it may face the difficulty of estimation degradation caused by the mismatch between the prior model sets and the real models. Based on the idea of variable structure and graph theory knowledge,this paper proposes a compensation variable structure interacting multiple model( CPVSIMM)algorithm by combining with covariance-matching technique. By analyzing the condition of estimation degradation to uniformly accelerated linear motion because of traditional VSIMM only selecting angular velocity as the model parameters,this proposed algorithm selects acceleration and angular velocity as the model parameters,and establishes digraph connectivity relationships and the principle of adapting model subset. Theoretical analysis and comparison of simulation results show that it can achieve higher fusion estimation accuracy,and reduce the cumulative error due to model sets mismatch and not timely switch.