Transformer-Based Reinforcement Learning for Autonomous Orbital Collision Avoidance in Partially Observable Environments
This work addresses collision avoidance for space operations, but it is incremental as it builds on existing POMDP methods with a transformer-based approach.
The paper tackles autonomous orbital collision avoidance in partially observable environments by introducing a Transformer-based Reinforcement Learning framework, which achieved more reliable operation under imperfect monitoring conditions compared to traditional architectures.
We introduce a Transformer-based Reinforcement Learning framework for autonomous orbital collision avoidance that explicitly models the effects of partial observability and imperfect monitoring in space operations. The framework combines a configurable encounter simulator, a distance-dependent observation model, and a sequential state estimator to represent uncertainty in relative motion. A central contribution of this work is the use of transformer-based Partially Observable Markov Decision Process (POMDP) architecture, which leverage long-range temporal attention to interpret noisy and intermittent observations more effectively than traditional architectures. This integration provides a foundation for training collision avoidance agents that can operate more reliably under imperfect monitoring environments.