Abstract Active debris removal (ADR) allows for the disposal of inactive satellites and larger objects, preventing the build-up of space junk and allowing to replace aging agents in a constellation. To make ADR missions more commercially viable, the removal and disposal of multiple debris objects using a single spacecraft are investigated. This paper proposes the use of artificial neural networks (ANNs) to quickly estimate the cost and duration of the transfers to de-orbit a range of debris objects, so that it is possible to identify the optimal sequence of objects which minimizes the cost and/or the duration of the mission, for the maximum number of de-orbited objects. To this end, the ANN is integrated within a sequence search algorithm based on a tree search. The performance of the proposed methodology is assessed by analyzing three distinctive sequences of multiple space debris removals. A near-term low-thrust propulsion technology enables to dispose of up to 13 debris objects within 10 years, when the optimal design parameters are chosen. The use of ANN allows for this solution to be found 26 times faster than current methods, while enabling the selection of faster and less expensive (being the propellant mass required lower) options.
Highlights Artificial Neural Networks (ANN) can be trained to estimate complex transfer models. ANN can quickly estimate transfers for the disposal of multiple space debris. ANN is used within a sequence search algorithm to identify optimal missions. Missions where up to 13 debris can be disposed of within 10 years are obtained. The use of machine learning greatly reduces the computational time by 26 times.
Design of multiple space debris removal missions using machine learning
Acta Astronautica ; 193 ; 277-286
2021-12-27
10 pages
Article (Journal)
Electronic Resource
English