With continuous developments in the space industry, the space near the Earth is occupied by a variety of spacecraft whose number is increasing dramatically every year. To avoid a collision, huge computation power is necessary to determine the possibility of a collision between two space objects. However, there are various uncertainties in the collision prediction process, which aggravates the burdens on space safety management.

Since the collision probability is usually applied to evaluate a dangerously close encounter, improving the precision of orbit prediction and covariance prediction is key.

In a research paper recently published in Space: Science & Technology, Zhaokui Wang, from Tsinghua University, proposed an efficient method with a back propagation (BP) neural network to improve the accuracy of orbit prediction and position error covariance prediction of space targets.

Wang's team also applied the proposed method to estimate the collision probability for the Q-Sat and space debris with NORAD ID of 49863.