For the problem of states estimation of nonlinear systems with respect to unknown parameters, a filter named interaction cubature Kalman filter(InCKF) is proposed. The novel filter is consisted of multiple concurrent CKFs interlacing with a maximum posteriori(MAP) estimator. By taking advantage of special properties of second order of Stirling’s interpolation and unscented transformation to approximate nonlinear functions, the unknown parameters are estimated and the performance of InCKF does not depend on the precision of model parameters. Furthermore, the states of system are estimated by using the cubature Kalman filter(CKF). The simulation results show that the InCKF is more accurate and stabilized than the classical method of state augmentation in the situation that model parameters are unknown.