TY - JOUR
T1 - Artificial Neural Network Based Mission Planning Mechanism for Spacecraft
AU - Li, Zhaoyu
AU - Xu, Rui
AU - Cui, Pingyuan
AU - Zhu, Shengying
N1 - Publisher Copyright:
© 2018, The Korean Society for Aeronautical & Space Sciences and Springer Nature Singapore Pte Ltd.
PY - 2018/3/1
Y1 - 2018/3/1
N2 - The ability to plan and react fast in dynamic space environments is central to intelligent behavior of spacecraft. For space and robotic applications, many planners have been used. But it is difficult to encode the domain knowledge and directly use existing techniques such as heuristic to improve the performance of the application systems. Therefore, regarding planning as an advanced control problem, this paper first proposes an autonomous mission planning and action selection mechanism through a multiple layer perceptron neural network approach to select actions in planning process and improve efficiency. To prove the availability and effectiveness, we use autonomous mission planning problems of the spacecraft, which is a sophisticated system with complex subsystems and constraints as an example. Simulation results have shown that artificial neural networks (ANNs) are usable for planning problems. Compared with the existing planning method in EUROPA, the mechanism using ANNs is more efficient and can guarantee stable performance. Therefore, the mechanism proposed in this paper is more suitable for planning problems of spacecraft that require real time and stability.
AB - The ability to plan and react fast in dynamic space environments is central to intelligent behavior of spacecraft. For space and robotic applications, many planners have been used. But it is difficult to encode the domain knowledge and directly use existing techniques such as heuristic to improve the performance of the application systems. Therefore, regarding planning as an advanced control problem, this paper first proposes an autonomous mission planning and action selection mechanism through a multiple layer perceptron neural network approach to select actions in planning process and improve efficiency. To prove the availability and effectiveness, we use autonomous mission planning problems of the spacecraft, which is a sophisticated system with complex subsystems and constraints as an example. Simulation results have shown that artificial neural networks (ANNs) are usable for planning problems. Compared with the existing planning method in EUROPA, the mechanism using ANNs is more efficient and can guarantee stable performance. Therefore, the mechanism proposed in this paper is more suitable for planning problems of spacecraft that require real time and stability.
KW - Artificial neural networks
KW - Autonomous control
KW - Deep space exploration
KW - Mission planning
UR - http://www.scopus.com/inward/record.url?scp=85045267992&partnerID=8YFLogxK
U2 - 10.1007/s42405-018-0006-6
DO - 10.1007/s42405-018-0006-6
M3 - Article
AN - SCOPUS:85045267992
SN - 2093-274X
VL - 19
SP - 111
EP - 119
JO - International Journal of Aeronautical and Space Sciences
JF - International Journal of Aeronautical and Space Sciences
IS - 1
ER -