Spacecraft autonomous rendezvous and docking guidance strategy generation method based on reinforcement learning

Abstract

The invention discloses a spacecraft autonomous rendezvous and docking guidance strategy generation method based on reinforcement learning. The method comprises the following steps: modeling a spacecraft rendezvous and docking process into a Markov decision process model; solving the Markov decision process model by adopting a dynamic programming algorithm to obtain scores of different actions adopted in all states, and generating a decision table; taking all states in the decision table as training data features, taking scores of all states in the decision table under each action as training data labels, and constructing training data; constructing a neural network model, and training the neural network model by adopting the training data to obtain a neural network model serving as approximate representation of the decision table; for a certain state, calculating scores of all actions in the state through the obtained neural network model, and selecting the action with the maximum score as the optimal guidance strategy; and based on the optimal guidance strategy, enabling the spacecraft to perform autonomous rendezvous and docking.

Description

technical field [0001] The invention belongs to the technical field of aerospace, and in particular relates to a method for generating a guidance strategy for autonomous rendezvous and docking of spacecraft based on reinforcement learning. Background technique [0002] With the increasing complexity of space missions, autonomous rendezvous and docking of spacecraft has become a challenging problem. The traditional solution is to use optimal control methods, which either make a lot of simplifying assumptions about the dynamic model, or require redundant computing resources. In recent years, reinforcement learning methods have been widely used in industrial applications such as robotic systems, autonomous vehicles, and the Internet of Things. Therefore, under the impetus of complex aerospace guidance tasks, some studies have introduced reinforcement learning techniques to enhance the autonomous rendezvous and docking guidance capabilities of spacecraft. A large number of exi...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: In order to realize the modeling of the autonomous rendezvous and docking process of spacecraft, to solve the problem of how to compress the memory volume required by the decision table, and to solve the problem of how to verify the effectiveness and robustness of the strategy table after compression, the present invention proposes A strategy generation method for spacecraft autonomous rendezvous and docking guidance based on reinforcement learning

Images