Autonomous artificial neural network star tracker for spacecraft attitude determination

An artificial neural network based autonomous star tracker prototype for precise spacecraft attitude determination is developed. Night sky testing is used to validate a system consisting of a charged-coupled-device-based camera head unit and integrated control hardware and software. The artificial neural network star pattern match algorithm utilizes a sub catalog of the SKY2000 star catalog. The experimental results are real time comparisons of the star tracker observed motion with the rotational motion of the Earth. The results of a field-programmable-gate-array-based implementation of the star pattern match algorithm are also presented.; A new technique of star pattern encoding that removes the star magnitude dependency is presented. The convex hull technique was developed in which the stars in the field of view are treated as a set of points. The convex hull of these points is found and stored as line segments and interior angles moving clockwise from the shortest segment. This technique does not depend on star magnitudes and allows a varying number of stars to be identified and used in calculating the attitude quaternion. This technique combined with feed-forward neural network pattern identification created a robust and fast technique for solving the “lost-in-space” problem.; The time required to solve the “lost-in-space” problem for this star tracker prototype is on average 9.5 seconds. This is an improvement over the 60 seconds needed by the current off-the-shelf autonomous star tracker by Ball Aerospace, the CT-633. Initial acquisition after launch as well as recovery from a loss of attitude knowledge during the mission would occur significantly faster with this prototype system when compared to current commercially available autonomous star trackers.