In order to satisfy the needs of Mars precision landing, a Mars entry navigation method is proposed for the problem of Mars atmospheric density model uncertainty. The navigation system processes accelerometer outputs as measurements, employs an extended Kalman filter bank regulated by the generalized multiple-model adaptive estimation method. Simulation results demonstrate the navigation system can identify the real atmospheric density model automatically, and show adaptivity and robustness to the uncertainty of atmospheric density. The navigation performance is greatly improved compared with traditional dead-reckoning