The agile earth observation satellite scheduling problem (AEOSSP) is a combinatorial optimization problem with time-dependent constraints. Recently, many construction heuristics and meta-heuristics have been proposed; however, existing methods cannot balance the requirements of efficiency and timeliness. In this paper, we propose a graph attention network-based decision neural network (GDNN) to solve the AEOSSP. Specifically, we first represent the task and time-dependent attitude transition constraints by a graph. We then describe the problem as a Markov decision process and perform feature engineering. On this basis, we design a GDNN to guide the construction of the solution sequence and train it with proximal policy optimization (PPO). Experimental results show that the proposed method outperforms construction heuristics at scheduling profit by at least 45%. The proposed method can also calculate the approximate profits of the state-of-the-art method with an error of less than 7% and reduce scheduling time markedly. Finally, we demonstrate the scalability of the proposed method.

The agile earth observation satellite scheduling problem (AEOSSP), as a time-dependent and arduous combinatorial optimization problem, has been intensively studied in the past decades. Many studies have proposed non-iterative heuristic construction algorithms and iterative meta-heuristic algorithms to solve this problem. However, the heuristic construction algorithms spend a relatively shorter time at the expense of solution quality, while the iterative meta-heuristic algorithms accomplish a high-quality solution with a lot of time. To overcome the shortcomings of these approaches and efficiently utilize the historical scheduling information and task characteristics, this paper introduces a new neural network model based on the deep reinforcement learning and heuristic algorithm (DRL-HA) to the AEOSSP and proposes an innovative non-iterative heuristic algorithm. The DRL-HA is composed of a heuristic construction neural network (HCNN) model and a task arrangement algorithm (TAA), where the HCNN aims to generate the task planning sequence and the TAA generates the final feasible scheduling order of tasks. In this study, the DRL-HA is examined with other heuristic algorithms by a series of experiments. The results demonstrate that the DRL-HA outperforms competitors and HCNN possesses outstanding generalization ability for different scenario sizes and task distributions. Furthermore, HCNN, when used for generating initial solutions of meta-heuristic algorithms, can achieve improved profits and accelerate interactions. Therefore, the DRL-HA algorithm is verified to be an effective method for solving AEOSSP. In this way, the high-profit and high-timeliness of agile satellite scheduling can be guaranteed, and the solution of AEOSSP is further explored and improved.

To prevent the proliferation of space debris and stabilize the space environment, active debris removal (ADR) has increasingly gained public concern. Considering the complexity of space operations and the viability of ADR missions, it would be necessary to schedule the ADR process in order to remove as much debris as possible. This paper presents an active debris removal mission planning problem, devoted to generate an optimal debris removal plan to guide the mission process. According to the problem characteristics, a two-layer time-dependent traveling salesman problem(TSP) mathematical model is established, involving the debris removal sequence planning and the transfer trajectory planning. Subsequently, two main novel methods based on machine learning are proposed for the ADR mission planning problem, including a deep neural networks(DNN)-based estimation method for approximating the optimal velocity increments of perturbed multiple-impulse rendezvous and an reinforcement learning(RL)-based method for optimizing the sequence of debris removal and rendezvous time. Experimental results of different simulation scenarios have verified the effectiveness and superiority of the proposed method, indicating the good performance for solving the active debris removal mission planning problem.

Simultaneously optimizing monetary cost and makespan of workflow execution is substantial to enhance the competitiveness of cloud services, but it still imposes challenges. Heuristics-based algorithms are often problem-dependent and well-suitable for special cases, but the complexity of workflow structures seriously challenges their generalization. Metaheuristics-based algorithms pose good generalization, but the elasticity and heterogeneity of cloud resources seriously impact their search efficiency. Inspired by previous works, we tailor a memetic workflow scheduling algorithm (KDMA for short) that embeds a heuristic local search operator into the multi-objective metaheuristic algorithm to combine their strengths and complement each other's shortcomings. Specifically, the proposed local search operator searches for a set of solutions with good convergence and diversity by accumulating tasks one by one. This operator is good at gathering workflow tasks into a limited range of candidate resources, thereby guiding the metaheuristic algorithm to focus on exploring potential solution regions. Moreover, KDMA includes an adaptive strategy to determine the number of solutions reproduced by the problem-special local search operator based on its past overall contribution. We compare KDMA to five representative algorithms over 25 real-world workflow instances to corroborate its superior all-around performance by performing the best on 21 workflow instances. Meanwhile, we conduct an ablation analysis to verify the performance contributions of two proposed mechanisms.