| With the development of space technologies,Mars explorations have become a hot research topic.From generation 1 represented by Mars Pathfinder to generation 2 represented by Curiosity,landing accuracy has been improved a lot from 150 km to 20 km to the target.In the future,to satisfy the scientific tasks,the 3rd generation landers should land accurately,namely “pinpoint landing”.The landing errors should be within 100 m to the target.The improvement of landing accuracy is benefited from advanced navigation,guidance and control techniques during Mars entry,descent and landing(EDL).Among the three phases,Mars entry is the most dangerous,challenging and important part.Advanced guidance and control are based on accurate state estimations from navigation system.So in this thesis,we study the innovative navigation scheme and advanced navigation algorithm during Mars entry,which has some uncertainties including atmosphere density,lift over drag and ballistic coefficient,as well as measurement outliers and some unknown disturbances like sudden storm.The main content of this thesis can be summarized as follows:1.The integrated navigation schemes have been investigated during Mars entry including IMU/orbiter based integrated navigation,and IMU/orbiter/Mars surface beacon(MSB)based integrated navigation.First,the condition number of observability matrix based observability analysis method is proposed to evaluate the navigation schemes.Then the estimability of the states is investigated when the navigation system is totally observable.By simulations,it is demonstrated that IMU/orbiter based integrated navigation is barely observable.Then IMU/orbiter/MSBs based integrated navigation scheme is investigated,which can not only provide more accurate state estimation,but also the navigation system is totally observable.2.The desensitized extended Kalman filter(DEKF)is proposed during Mars entry under atmosphere density uncertainty,lift to drag uncertainty,and ballistic coefficient uncertainty.The differences between DEKF and standard EKF are in the calculation of the gain matrix.The gain matrix from DEKF are obtained by minimizing a cost function that is composed of the trace of posterior covariance matrix and the weighted norm of the posterior state estimation error sensitivities(PSEES).The DEKF is robust with unknown disturbances and uncertainties.By applying IMU/Mars orbiters/MSBs integrated navigation,simulations demonstrate that the introduced DEKF is far less sensitive to uncertain parameters than standard EKF during Mars entry.With the increase of uncertainties one by one,the RMSE of the states are still convergent in case of DEKF as compared to standard EKF where they become larger.At last,the consistency test is carried on to further validate the proposed DEKF.3.A novel version of adaptive divided difference filter(DDF)is proposed under uncertainties or unknown biases for the complicated Mars entry dynamic and measurement models.This new kind of DDF is founded on the principle of covariance matching.It introduces adaptive forgetting factor to amplify the predictive estimation error covariance.By this way,the adverse effect from dynamic model errors or uncertain parameters can be compensated.To make the states can be estimated by different channels and different rates,the scalar forgetting factor is further extended to multiple forgetting factors in terms of diagonal matrix.By Monte Carlo simulations,it is verified that the proposed DDF has more accurate state estimations than general DDF,and has better credibility under uncertainties.Under unknown biases or disturbances,the estimation accuracy of proposed DDF is far better than general DDF,and the forgetting factors can vary adaptively according to the varying of unknown biases or disturbances,which verified the adaptive capability of the proposed DDF.4.The Huber DDF and adaptive Huber DDF are proposed to solve the problem of biases existing in the state model and measurement model during Mars entry.First,based on the generalized maximum likelihood perspective,the solution of the standard Kalman filter is derived by minimizing a cost function.Under tha case of biased state model and measurement model,the cost function of the standard Kalman filter is modified by the robust Huber function,and a new cost function are obtained,based on which new predictive state estimation error covariance and measurement noise covariance are derived.By embedding the new predictive state estimation error covariance and measurement noise covariance into the frame of DDF,the Huber DDF is derived.Then on the basis of Huber DDF,its predictive state estimation error covariance are modified by introducing an adaptive forgetting factor,the adaptive Huber DDF is derived.By simulations,it can be demonstrated that the proposed adaptive Huber DDF has more accurate state estimations than Huber based DDF,and is much more accurate than general DDF.Further more,the forgetting factor in the proposed adaptive Huber DDF is smaller than that of the adaptive DDF proposed above.5.The uncertain parameters exist in Mars entry have a significant effect on Mars entry navigation.In this paper,the hierarchical adaptive filters regulated by a gating network is presented to identify the uncertain parameters during Mars entry,which are atmosphere density,lift over drag ratio and ballistic coefficient.The hierarchical structure is composed of several bank of filters,and each bank of filters is composed of several experts running in parallel.Each expert is a realization of some parameter values.The gating network can adaptively assign appropriate weights to the filter banks and the experts.The weight which is close to unity corresponding to certain expert means that the given parameter value is close to the real value or is right the real value.By this way,the uncertain parameters can be identified.Also the effect of learning rate on the convergence rate of gating weights and the identification capability is demonstrated.In the end,main results in this thesis are summarized,and the prospects for the future research are presented. |
PDF Full Text Request