Abstract Maintaining tracks on space objects with limited sets of observers is a critical problem, made more urgent with exponential growth in the population of near-Earth satellites. An optimally convergent decision making methodology is proposed for sensor tasking, using the Monte Carlo Tree Search methodology. This methodology is underpinned by the partially observable Markov decision process framework; it utilizes polynomial exploration of the action space, and double progressive widening to avoid curses of history. The developed tasking techniques are applied to a large-scale application considering the tracking problem in the emerging cislunar regime. Uncertainty studies are performed for a set of 500 objects in a variety of candidate periodic and highly elliptical orbits, with realistic sensor models incorporating physical parameters and explicit probability of detection. These simulations are utilized as a means to evaluate observer quality, considering candidate space-based sensors following L1 Lyapunov and L2 Northern Halo orbits. Results demonstrate the importance of space-based observers for maintaining estimates on objects in cislunar space and give insight into the criticality of relative motion between observers and targets when optical measurements are utilized.
Sensor tasking in the cislunar regime using Monte Carlo Tree Search
Advances in Space Research ; 70 , 3 ; 792-811
2022-05-02
20 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Optimal Cislunar Architecture Design Using Monte Carlo Tree Search Methods
Springer Verlag | 2023
|Optimal Cislunar Architecture Design Using Monte Carlo Tree Search Methods
British Library Conference Proceedings | 2022
|Monte Carlo Tree Search Methods for Telescope Tasking
AIAA | 2020
|MONTE CARLO TREE SEARCH METHODS FOR TELESCOPE TASKING
TIBKAT | 2020
|Cislunar Space Situational Awareness Sensor Tasking using Deep Reinforcement Learning Agents
British Library Conference Proceedings | 2022
|