"Fusion on Autonomous Navigation Based on Unscented Kalman Filter"

Shu-Lin Sui, Jian Yuan, and Wen-Xia Zhang
Qingdao University of Science &Technology


Through much research into environment characteristic of the fast movement and anomalous celestial bodies figure in autonomous orbit determination of deep-space explorer, an autonomous optical navigation scheme based on Unscented Kalman Filter is proposed to determine the circling orbit state in time, in which star sensor and navigation camera are used in measuring special angles between planetoids and explorer, which can promote parameter mean and predictive precision, so it can solve the difficulty of autonomous orbit determination. Compared with extended kalman filter, it shows that its performance is better than extended kalman filter based on Minimus Mean Square Error.

The shape of planetoid is irregular, and its quality is non-uniform and the perturbation caused by solar radiation pressure and other celestial bodies and so on, environment of explorers circling round small celestial bodies changes very quickly, moreover, the explorer's track dynamics model is not easy to determine [1, 2], which sets a higher request to the precision and robustness of guidance system.

The traditional EKF method, because of the partial linearization to transfer function and the measure function, can not adapt the serious non-linear situation; moreover, derivate matrix of nonlinear function can not be carried out by using the partial derivative. On the other hand, it is every difficult for us to  survey directly to the detector position and the speed for the autonomous navigation, then it is necessary to survey that by using the track dynamics model and a series of measures to estimate explorer track. The choice of conditional model includes the choice of conditional vectors as well as determination on scale of state equation. The choice of observation model includes the selection of observation parameter, as well as computation model based on observation parameter establishment and observation simulation. The observation model and the conditional model must be consistent.

One orbit-determining method of autonomous optical navigation method based on Unscented Kalman Filter is proposed in this paper. First, using the small celestial body image and the referenced model image of small celestial body captured by the navigation camera, to determine the orientation of explorer relatively to the small celestial body and of edge characteristic points , combined the shape model of small celestial body and the posture information about explorer, the definite position vector can be calculated, then which can be input to the Unscented kalman filtering algorithm to independently determines the small celestial body track.

2006 MAPLD International Conference Home Page