Investigations Of Autonomous Orbit Determination Based On Ground Remote-Sensing Images

The development of aerospace exploration has raised higher demands for satellite orbit determination(OD)capabilities.With more man-made satellites being launched,traditional determination method based on ground stations are faced with the challenges of higher operational costs and poorer survivability.Therefore,it is of great importance to develop autonomous satellite OD strategies for future space projects.One example is the ground image-based autonomous OD strategy,which has a good application prospect due to the easy acquisition of observational data,high-precision measurement and abundant information contained in the imagery.Given this background,the paper focuses on OD schemes using optical imagery of ground objects and serves to expand current knowledge on this topic.Three different OD schemes are proposed and their performances were evaluated based on Monte-Carlo simulations.The main results are as follows:1.A new autonomous OD scheme based on images of ground point objects is proposed and investigated with Monto-Carlo simulations.Based on the perspective projection,a measurement model which uses focal plane coordinates as measurement is first established.With the help of extended Kalman filter(EKF)or Unscented Kalman filter(UKF),the complete scheme can then be constructed.Comparison of the scheme performance using the two filters shows similar results.The obtained root-mean-square(RMS)errors of position and velocity are approximate 10 m and stay below 2 cm s-1,respectively.The influence of lighting constraints(night-time and cloud coverage constraints),image resolution and pointing accuracy on the OD performance are also investigated.Results demonstrate the greater impact of pointing accuracy.2.A new orbit determination scheme targeting at communication and remote sensing satellites in a hybrid constellation is proposed by using of combination of ground imagery data,altimeter data and inter-satellite data.We first designed one such hybrid constellation with a two-layer configuration(LEO/MEO)by optimizing coverage and revisit cycle.Based on the hybrid constellation,the OD performance of the scheme using different combinations of measurement data is investigated with respect to individual remote-sensing satellite and the constellation.For individual satellite in the constellation,the combined use of altimeter data pointing to geometer calibration sites and imagery improves performance compared to that using only imagery,while the combined use of ocean altimeter data with imagery results in a poorer performance.For the constellation,when three measurement datasets are used,orbits of all the satellites in the hybrid constellation can be autonomously determined and the orbit determination precision for LEO satellites is further improved.Our results also show that the performance of the scheme would be affected when remote sensing observations on certain satellites are absent.The orbit determination performance is also investigated in real situations where remote sensing observations on certain satellites are absent.Results show no difference in the orbit determination precision between satellites with and without a lack of remote-sensing observations.However,the orbit precision of theconstellation declines when compared to the ideal situations.3.Current image-based orbit and attitude determination(OAD)schemes mainly use the light-of-sight(LOS)vectors to ground point features as measurement and have two drawbacks.First,valid ground point features might not always be available.Besides,regarding large-size landmarks that cover several pixels in the image as point features could affect the precision of the LOS vectors.Aimed at this problem,we present a new OAD scheme using onboard images of regular-shaped ground objects.The large availability of regular-shaped objects in man-made constructions provide a feasibility for the new scheme.Firstly,derivation processes of the pose estimation algorithms using parallelogram,orthogonal corners,ellipse and ellipsoid are provided.The required shape properties to obtain different output in the four algorithms are summarized.Based on this,two OAD schemes are proposed.When only shape constraints of the object shapes are known,the pose estimation algorithms give solution to the LOS vector,which can be used as measurement.Then an OAD scheme consisting of a dynamic model,a measurement model and an EKF filter can be established.Results from Monte-Carlo simulations demonstrate the feasibility of the scheme and show the substantial influence from both lighting constraints and image sampling frequency on its OAD performance.For parallelogram and orthogonal corners,when additional size properties are known,position and attitude parameters can be directly obtained purely based on the pose estimation algorithms,which means the position and attitude of the satellites can be determined directly.This scheme is referred to as the deterministic scheme.Comparison of OAD performance between two schemes shows a better performance of the first scheme.