Relative Navigation of Multiple Spacecraft during Proximity Maneuvers

3 I. INTRODUCTION A number of d ifferent a pproaches to t he m aneuvering t arget p roblem have appeared in recent l iterature [ 5].[8],[12],[14],[17],[18],[19],[20] a nd [21]. Modeling a system a ccurately, w hen u sing a K alman f ilter f or t arget t racking, i s on e of the m ost important pr oblems. If the s ystem model is not accurate, track loss may occur quickly. The d evelopment of a n accurate s ystem m odel r equires m aneuvers detection a nd estimation of the magnitude of the maneuvers [5], [8], [17] and [20]. In space applications, particularly in spacecraft relative navigation for spacecraft autonomous r endezvous a nd assembly, each spacecraft i s t he t arget of a nother one . In other words, there is a complete symmetry, where each vehicle needs to track the others. Here a n a dditional c hallenge i s e ncountered: the f requent loss of communication and, when the application involves more than one spacecraft, the contact for the data exchange could not be always available. If data communication is not the way spacecraft perform relative n avigation, a vision b ased s ystem may be used. These types of sensors require image processing and may result in low frequency measurement upda tes, especially for small spacecraft w ith l imited computation capabilities. Such sensors suffer of p roblems such a s l imitations on t he f ield of view a nd/or ot her s pacecraft obs tructing t he view. Furthermore, in spacecraft relative maneuvering, each vehicle does not usually know the other vehicles inputs, i.e. it does not possess the information about the maneuvers decided by its fellow spacecraft. This missing information needs to be reconstructed somehow in the Kalman filter, which otherwise would diverge quickly. Many schemes addressing these types of problems have been developed, based on different a ssumptions about t he t arget’s dynamics. For e xample, some r esearchers developed algorithms based on constant velocity o r c onstant a cceleration assumptions, other methods need to use small s ampling times between two measurements to operate accurately (Munu e t a l., 1992 ; B lair e t a l., 1991; Mook and Shyu, 1992; Rokhsaz and Steck, 1991). The above listed limitations are the motivation for the present study.