Autonomous proximity operations have recently become appealing as space missions. In particular, the estimation of the relative states and inertia properties of a noncooperative spacecraft is an important but challenging problem, because there might be poor priori information about the target. Using only stereovision measurements, this study developed an adaptive unscented Kalman filter to estimate the relative states and moment-of-inertia ratios of a noncooperative spacecraft. Because the accuracy of the initial relative states has an effect on the estimation convergence performance, attention was also given to their determination. The target’s body-fixed frame was defined in parallel to the chaser’s initial body-fixed frame, and then the initial relative attitude was known. After formulating kinematic constraint equations between the relative states and multiple points on the target surface, particle swarm optimization was utilized to determine the initial relative angular velocity. The initial relative position was also determined under the assumption that the initial relative translational velocity was known. To estimate the relative states and moment-of-inertia ratios using the adaptive unscented Kalman filter, the relative attitude dynamic model was reformulated by designing a novel transition rule with five moment-of-inertia ratios, described in the defined target’s body-fixed frame. The moment-of-inertia ratios were added to the state space, and a new state equation with variant process noise covariance matrix Q was formulated. The measurement updating errors of the relative states were utilized to adaptively modify Q so that the filter could estimate the relative states and moment-of-inertia ratios in two stages. Numerical simulations of the adaptive unscented Kalman filter with unknown moment-of-inertia ratios and the standard unscented Kalman filter with known moment-of-inertia ratios were conducted to illustrate the performance of the adaptive unscented Kalman filter. The obtained results showed the satisfactory convergence of the estimation errors of both the relative states and moment-of-inertia ratios with high accuracy.
Vehicle Driving State Estimation Using an Improved Adaptive Unscented Kalman Filter
