Learning-based framework for multistage space debris collision warning and avoidance

doi:10.1016/j.ast.2026.112305

Learning-based framework for multistage space debris collision warning and avoidance

2026-04-06

0

PRE

AI

DOI：10.1016/j.ast.2026.112305

Original

Original request for help

Save

Abstract

En 中文

• A progressive multistage collision scenario framework is proposed to simulate the evolving risks of encountering new debris after initial avoidance maneuvers. • A unified comparison of the three learning-based methods, MCTS, CEM and PPO, provides practical insights into the scalability and stability required for large-scale orbital safety. • A two-stage early warning screening mechanism is integrated into the learning-based multistage collision avoidance framework, improving computational efficiency and scalability. • The proposed autonomous framework successfully manages complex multistage scenarios involving over 100 debris objects from real and probabilistic sources.

Keywords:

multistage collision avoidance

space debris

machine learning

autonomous systems

orbital safety

Learning-based framework for multistage space debris collision warning and avoidance

Abstract

Journal

Organization