A new data fusion filter for rendezvous navigation with global positioning system aiding

A data fusion filter is designed and tested which optimally exploits separate, decentralized inertial and relative navigation filters without modification for accurate and robust rendezvous navigation. The filter uses a Global Positioning System (GPS) navigation filter, along with a suboptimal rendezvous filter used by NASA's human spaceflight programs for several decades. The fusion filter does not require GPS for the target vehicle. The filter is general enough to accommodate common and unique elements in the decentralized filters' states, and its estimates may be periodically fed back to reinitialize one or both of the decentralized filters, enhancing their performance. It is proved that a necessary condition for fusion after feedback is that the rank of a matrix with the filters' observability grammians along its diagonal be at least equal to the dimension of the common state elements. The fusion filter accommodates a suboptimal perfect target assumption in the decentralized rendezvous filter by exploiting a duality of information in the rendezvous filter's chaser and relative state estimates.;The fusion filter shows promising performance in simulations of Space Shuttle rendezvous, and when it is applied to data from a Space Shuttle mission. Critical to the success of the flight data processing are a fusion edit test based on the state vector difference of the two filters, and underweighting of the rendezvous filter when it is clear that large inertial biases cause failure of the edit test. The results indicate that the fusion filter is a viable enhancement to existing rendezvous navigation tools, as well as an alternative to new development of integrated inertial/relative rendezvous filtering approaches, allowing maximal use of and minimal impact to existing filters in an integrated rendezvous navigation solution.