Research On Multi-scheme Navigation Methods For Lunar Landing

With the successful landing of the Chang'e-4 on the Lunar far-side and the successful launch of the Chang'e-5,the"Orbit-Fall-Back"three-step strategy of Chinese Lunar Exploration Program has entered the sample return stage.As the core component of the Lunar lander,the pose estimation accuracy of the navigation system is crucial for achieving a precise and controlled soft landing in high-value areas on the Lunar surface.Compared to the attitude,the acquisition of high-precision position of the lander is more difficult due to the large signal transmission delay between the Earth and the Moon,the lack of a priori information about the Moon,the limited availability of auxiliary information,etc.This paper takes Lunar landing as the research background,and mainly focuses on the multi-scheme navigation and localization methods,which include:According to the requirement of navigation system for lunar landing,the configu-ration scheme of multi-sensor integrated navigation system is designed,and the mech-anization equations of navigation system in different reference coordinate systems and the measurement models of IMU,monocular camera,radio beacon,laser altimeter,etc.,are established.For different mission characteristics of fully autonomous navigation and lunar-based aided navigation,the simulated landing trajectory was designed based on the parameters of Chang'e-4 and ATON probes respectively,and the IMU pose error under different trajectories was compared to verify the necessity of the research on multi-source fusion localization algorithm and lays the foundation for subsequent researches.To solve the problem that existing visual/inertial navigation systems cannot work throughout the landing descent process,which is induced by the camera field of view and crater distribution,a terrain-free visual/inertial relative navigation method is investigated.By introducing multi-scale image pyramids and image edge histograms,the time con-suming computation of the matching based tracking algorithm and the poor robustness of the optical-flow based tracking algorithm under large scale movements are solved.In addition,the cost function for visual inertial joint estimation is designed by combining the IMU pre-integration residuals and visual re-projection residuals,and the pose estimation framework is extended from the single-state recursive filtering to the multi-state sliding window filtering.The indoor equivalent test results show that the visual/inertial relative navigation system can achieve a reasonable relative position estimation accuracy.Although visual/inertial relative navigation system can achieve better relative position estimation accuracy,they cannot produce an effective absolute position estimation of the lander,which need to be aligned periodically by introducing absolute measurements.To address this issue,a new Sigma-point Kalman filter based on stochastic cloning framework(SC-SPKF)is investigated and extended to alternating asynchronous measurements to achieve an effective fusion of relative and absolute measurements.In addition,a robust adaptive SC-SPKF(RA-SC-SPKF)is further proposed by combining the covariance matching and multiple fading?~2test methods with the SC-SPKF in order to suppress the time-varying measurement noise caused by elevation changes during landing descent,as well as the outlier caused by terrain changes and crater mismatches.The simulation results show that RA-SC-SPKF achieves good position and velocity estimation accuracy under various interference conditions.On the other hand,with the proposal of constructing a Lunar residential station,Lunar-based supported navigation methods have become a hot research direction in the navigation system design for Lunar landing.As a typical representative,the Lunar-based supported navigation based on radio beacons require an exactly priori knowledge of beacon locations,and since there isn't any reliable precise positioning system on the Moon,the position determination accuracy based on ground-based deep space networks or lunar orbiters is usually in the order of hundred meters.To solve the problem of lander navigation and positioning in this case,a distributed beacon supported navigation method based on sparse extended hybrid filtering(SEHF)is designed to estimate both lander and beacon positions simultaneously.On this basis,a Robust Adaptive Iterated SEHF(RAISEHF)algorithm is proposed by combining the damping adaptive iterative algorithm and the robust Triggs correction algorithm.The simulation results show that the proposed RAISEHF can effectively suppress the effects of model linearization errors and non-Gaussian measurement noise on the filter estimation accuracy.