Study On Optical Measurement Based Relative Navigation And Its Application In Planetary Landing

Autonomous landing probe on planetary surface and achieving pinpoint planetary landing is a rather challenging task. Probe must possess the ability of accurate relative navigation. Traditional dead-reckoning based navigation mode can't meet the accuracy requirement of future planetary pin-point landing mission. So it's necessary to develop the new generation optical measurement based planetary landing relative navigation algorithms.With the supports of the Tenth Five-Year 863 Program'Autonomy Technology of Deep Space Exploration and its Simulation and Demonstration System', this dissertation deeply and systemically studies the optical measurement based relative navigation algorithm and its application in planetary pin-point landing from the following two aspects: optical measurement relative navigation algorithm and soft landing asteroids autonomous navigation semi-physical simulation. The main contents of this dissertation are as follows.Firstly, vector measurement based autonomous relative navigation algorithm is proposed, which is based on vector measurement between the probe and the feature points on the surface of target celestial body. The position vectors from the probe to three non-collinear feature points can be constructed form the measurements of optical navigation camera and laser range finder, and then, the feature point fixed coordinate system and the probe relative position and attitude, defined in the landing site coordinate system, can be directly constructed. Feature points tracking and inheritance based autonomous optical measurement relative navigation algorithm is developed to overcome the problem of navigation failure due to feature points outflow from camera field of view.Secondly, line-of-sight (LOS) measurement based autonomous relative navigation algorithm is presented, which is based on LOS measurement between the probe and the feature points on the surface of target celestial body. The six degree-of-freedom motion parameters of the probe, relative position and attitude, can be uniquely determined using nonlinear estimation algorithm (batch method or kalman filter) processing four or more than four LOS observations. In order to overcome Gaussian Least Square Differential Correction algorithm convergence difficulty when an accurate initial estimate is not given, LOS measurement relative navigation Levenberg-Marquardt algorithm is developed.Next, observability analysis of LOS measurement relative navigation system is deeply analyzed. Navigation observation model is constructed based on collinearity equations, optimal error variance matrix and Fisher information matrix, satisfying Cramer-Rao lower bound, are obtained using maximum likelihood estimation. Observability and observable degree of LOS measurement relative navigation system are studied by analyzing the rank, eigenvalue and eigenvector of Fisher information matrix. Theory analysis above-mentioned is confirmed by Matlab symbolic operation.Then, optical measurement aided inertial navigation algorithm is developed to overcome the shortcoming of optical navigation and inertial naviagiton alone. Natural feature points are identified and tracked by use of optical navigation camera and onboard software; and then, the landmark image information derived from navigation camera and the spacecraft state information sensed by IMU (Inertial Measurement Unit) are integrated in extended kalman filter algorithm to effectively correct the constant biases of IMU and avoid the navigation interregnum due to tracked landmark drop-out.Lastly, combining with the Tenth Five-Year 863 Program'Autonomy Technology of Deep Space Exploration and its Simulation and Demonstration System', semi-physical simulation system of asteroid soft landing autonomous navigation is built up based on optical navigation camera, asteroid geomorphy simulator and Matlab/Simulink/dSPACE integration simulation flatform. The feasibility of the proposed optical measurement based relative navigation algorithm is confirmed by semi-physical simulation and analysis.