An improved Monte Carlo tree search algorithm with dynamic classification for multi-spacecraft multi-target allocation

This paper proposes a solution framework based on the Monte Carlo Tree Search algorithm for the multi-spacecraft multi-target allocation problem in the active debris removal mission. The target allocation problem is modeled as a multi-step optimal decision-making problem by the Markov Decision Process. An improved Monte Carlo Tree Search algorithm is proposed to tackle the allocation problem to obtain the optimal target sequence. In the expansion operation, a dynamic classification mechanism is introduced to retain more valuable actions based on the orbital elements of targets. In the backpropagation and simulation operations, fuel consumption is employed as the value assessment criterion to reconstruct the objective of the algorithm. In the selection operation, the roulette wheel selection is applied to enhance the search efficiency. Numerical simulations are conducted to compare the improved Monte Carlo Tree Search algorithm with other commonly used algorithms. The results demonstrate that the proposed algorithm exhibits superior optimization capabilities (about 13 % fuel reduction in 40 target scenarios) and robustness compared to the commonly used algorithms, along with a more stable fuel-saving performance as the number of targets increases.