Research On Vision-aided Inertial Navigation And Its Application In Planetery Landing

Precision autonomous navigation is one of the key technologies to achieve theprecision planetary landing. The traditional dead-reckoning navigation method hasbeen proved unable to meet the accuracy requirement of the mission. In order toachieve precision landing on a remote planet surface, there is necessary to developnew types of navigation methods based on vision measurement.According to the demands for interplanetary landing mission, and thecharacteristics of different flight phases, in this paper, we studied the relevantnavigation methods using different kinds of visual measurements. The main contentof this paper are as follows:Firstly, the navigation method based on mapped landmarks is studid. In thedescent phase, we can extracts2D-to-3D correspondences between descent imagesand a surface map, and then get the position of the navigation feature points. Adinamic model of the spacecraft is builted, then the measurement model is given. Inorder to fuse the IMU measument with the mapped landmark measument, a filterwas designed to suppress the influence of IMU bias and system noise. Series ofthorough analysis were performed to determine the conjunctions between systemparameters with navigation performance.Secondly, the navigation algorithm using opportunistic features is proposed.With the decline of the spacecraft, the resolution difference between the descentimage and map will become larger and larger, which makes the mapped landmarksunable to use. At this situation, we can use the features tracked through a sequenceof descent images (opportunistic features). A measurement model which candescribe constraints between multiple camera poses was established. In order toobtain the precision estimation of spacecraft states, a corresponding filter wasdesigned to fusion the vision measurement with the output of IMU. And then, theperformance of the proposed navigation algorithm was validated by numericalsimulation.Then, the relative navigation method for the landing phase is studied. In thisphase, the altitude of the lander is in the measurement range of laser rangefinder.The observation vectors were costructed, using the measurement of laserrangefinder and camera. Based on the geometric relationship between the rangingvector and the landing point, a straightforward and practical navigation method wasdesigned to compute the relative position and attitude of the lander. A series ofanalysis was performed to determine the influences of correlative system parameters on navigation performance.