Research On Tracking Of Maneuvering Spatial Target

Space-based spacecraft tracking is one of essential issues on aerospace information processing and control, and plays an important role in successful implementation of space surveillance, autonomous rendezvous, satellite formation, on-orbit service and so on. The aim of space object tracking is to provide reliable real-time dynamic state of target, so as to support decisions made on target recognition, classification, catalogue, et al. However, because target's ability of non-cooperative maneuvering is increased and some space-based measurement sensors break down sometimes, space target tracking becomes difficult and its fault-tolerant is much harder to be guaranteed. Although many meaningful results in the problem of space target tracking have been obtained, it is still lack a space target autonomous tracking method with accuracy and fault-tolerant for the case that target exists non-cooperative orbital maneuvering and parts of on-board sensors have fault.For requirement of accuracy and fault-tolerant in tracking of space non-cooperative maneuvering target, space-based single sensor autonomous accurate tracking method and space-based multi-sensor information fusion fault-tolerant tracking method are mainly studied in this dissertation. In the background of that target maneuvers uncooperatively and that part of sensors have fault, the research is developed around four main aspects: relative motion model, single sensor and Gaussian noise based robust tracking algorithm, single sensor and Glint noise based robust tracking algorithm and multi-sensor information fusion fault tolerance tracking algorithm. The primary contributions of this paper are summarized as follow:1. Considering space target that probably maneuvers non-cooperatively under single sensor with Gaussian measurement noise, we propose a robust tracking algorithm: redundant adaptive robust unscented Kalman filter (RARUKF). RARUKF propagates nonlinear mean and variance by using unscented transform technique. It overcomes large truncation error and calculation difficult of Jacobian matrix caused by linearization technique in traditional RAREKF. At the same redundancy degree, RARUKF not only significantly reduce frequency of switching to robust filter status, but also make compensation magnitude in robust mode decreased, so that it reaches second-order accuracy in general. Hence, ability of resisting disturbance is enhanced and optimality of tracking result is improved. By introducing redundancy factor, RARUKF recovers adaptive switching mechanism of filtering status, which is failure in the proposed ARUKF when model error exists all the time and external disturbance comes forth randomly. In RARUKF, normal switching between robust and optimal filtering status is recovered, on the other hand, disturbance restraint becomes controllable, and disturbance tolerance degree can be set according to actual accuracy requirement, so that inner and external disturbance exceeds the redundancy range can be eliminated. Also, sufficient conditions for stability are developed, and the convergence of algorithm is proved.2. Considering space target that has probably non-cooperative maneuvering under single sensor with glint measurement noise, we propose a modified particle filter: redundant adaptive robust unscented particle filter. RARUPF use RARUKF to generate importance density function, so that the problem of serious deviation between true posterior probability density function and importance density function generated by UKF in traditional unscented particle filter (UPF) is resolved reasonably. RARUPF not only utilizes the latest measurement, but takes into account both model error and external disturbance, so that the effects of particle degeneracy are reduced, and makes importance density function more close to true posterior distribution. Hence, the tracking accuracy of the filtering algorithm is advanced. RARUPF integrates advantages come from robust filter, which can deal with inner disturbance and external disturbance, and particle filter, which can deal with glint noise, so it has robustness and optimality.3. Under the condition that parts of sensors probably break down, a type of multi-sensor information fusion fault tolerance tracking approach is proposed for space non-cooperative target. That is federal filter algorithm based on transient relative motion model, RARUKF robust tracking algorithm and residual detection method. Multi-sensor can provide measurement have redundancy, so basic information requirement of fault-tolerant tracking is satisfied. Transient tracking model can provide accurate and long-term description of relative orbital motion between two spacecrafts under target maneuvers and does not maneuver, so basic model requirement of fault-tolerant tracking is fulfilled. RARUKF, as filtering algorithm of fault-tolerant tracking, can provide accurate estimation of relative dynamic states of target. Residual detection method, as fault detection and isolation algorithm of fault-tolerant tracking, can detect timely the sensor malfunction and isolate the corresponding local estimate value. In federal filter structure design, main filter fuses local estimate values with no-reset mode. The steady global estimate value can be guaranteed in case of some sensors come forth fault, recovery from failure and system reconfiguration. Fault-tolerant and reliability of target tracking are ensured, such that the integrated function of multi-sensor information fusion fault-tolerant tracking can be realized and work stably.