| 摘 要: 为应对低地球轨道下潜在的航天器脉冲式轨道转移任务挑战,提出一种用深度强化学习算法建立轨道转移通用控制模型的方法,以减少人工干预,解决反应不及时等问题。通过对轨道动力学的建模和对马尔可夫决策过程的设计,成功将TD3(Twin Delayed Deep Deterministic Policy Gradient)算法运用于轨道转移决策,实现高度自主的脉冲式点火控制器的设计。实验结果表明,使用TD3算法建立的脉冲式点火控制器,在不同的轨道转移任务下自主到达目标轨道的成功率可达96.1%,同时完成了轨道5个根数的收敛,证明TD3算法用于解决该问题的可行性与有效性。 |
| 关键词: 轨道转移;深度强化学习;TD3算法 |
|
中图分类号: TP183
文献标识码: A
|
|
| General Orbit Transfer Control Based on TD3 Algorithm in Low Earth Orbit |
|
CAO Haitao, QIU Pengpeng, CAI Xia
|
(School of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, China)
caohaitao97@163.com; qpp9616@163.com; cxdaisy@zstu.edu.cn
|
| Abstract: To address potential impulsive orbit transfer tasks for spacecraft in low earth orbit, this paper proposes a method to establish a general control model for orbit transfer using deep reinforcement learning. This method aims to reduce human intervention and address issues related to delayed responses. By modeling orbital dynamics and designing a Markov decision process, the TD3 ( Twin Delayed Deep Deterministic Policy Gradient) algorithm is successfully applied to transfer decisions, achieving highly autonomous design of impulse ignition controllers. Experimental results demonstrate that the impulse ignition controller established with the TD3 algorithm can autonomously reach the target orbit in various orbit transfer tasks, with a success rate of up to 96. 1% , and the convergence of the five orbital parameters is achieved simultaneously, demonstrating the feasibility and effectiveness of the TD3 algorithm in addressing this problem. |
| Keywords: orbit transfer; deep reinforcement learning; TD3 algorithm |
