Memory-Efficient Meta-Reinforcement Learning for Adaptive Safety-Critical Control in Adversarial Spacecraft Proximity Operations
A new research paper explores the effectiveness of various recurrent neural network architectures and reinforcement learning algorithms for adaptive safety-critical control in spacecraft proximity operations. The study specifically compares Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), and Selective State Space Model (Mamba) networks, alongside Proximal Policy Optimization (PPO) and Soft Actor Critic (SAC) training algorithms. Results show that Mamba, when paired with PPO, demonstrated superior performance in task completion, safety, and fuel efficiency, even in adversarial scenarios. AI
IMPACT Demonstrates potential for advanced AI control systems in safety-critical aerospace applications.