Memory-Efficient Meta-Reinforcement Learning for Adaptive Safety-Critical Control in Adversarial Spacecraft Proximity Operations

2026-06-15

AI Summary

This work addresses the challenges of safety and fuel efficiency in spacecraft rendezvous and proximity operations under thrust constraints and adversarial target behavior. To this end, the authors propose a novel approach that integrates input-constrained control barrier functions (ICCBFs) with meta-reinforcement learning, enabling non-myopic and robust safety-critical control by learning the class-K functions that define the ICCBF recursion. The study presents the first systematic evaluation of memory-efficient architectures—including LSTM, GRU, and the state-space model Mamba—combined with PPO and SAC algorithms across both cooperative and adversarial scenarios. Experimental results demonstrate that the Mamba-PPO configuration significantly outperforms other methods in terms of mission success rate, safety, and fuel efficiency, thereby highlighting the advantages of state-space models for this class of control tasks.

Abstract

Autonomous spacecraft rendezvous and proximity operations (RPO) require controllers that guarantee safety under thrust constraints while minimizing fuel expenditure. Input-constrained control barrier functions (ICCBFs) provide a control method for nonlinear systems with actuation constraints that construct a forward-invariant safe set. Previous work has shown that learning class-$\mathcal{K}$ functions defining the ICCBF recursion via meta reinforcement learning (meta-RL) yields a robust, non-greedy approach to safety-critical control in RPO. This paper extends that framework further by investigating the performance of three recurrent network architectures (Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Selective State Space Model (Mamba)) and two training algorithms (Proximal Policy Optimization (PPO) and Soft Actor Critic (SAC)) to identify the best setup for tuning ICCBF class-K functions via meta-RL. In addition to cooperative test cases, performance is evaluated in the presence of adversarial behavior where the target spacecraft behaves in a way that worsens the safety of the chaser spacecraft. Results indicate that state space models such as Mamba when used with PPO achieve superior task completion, safety, and fuel-savings compared to other architectures, across all cooperative and uncooperative scenarios tested.

Problem

• spacecraft proximity operations
• safety-critical control
• adversarial behavior
• input-constrained control
• fuel efficiency

Innovation

• Meta-Reinforcement Learning
• Control Barrier Functions
• State Space Models
• Adversarial Spacecraft RPO
• Memory-Efficient Architecture