Risk-Sensitive Mission Planning via Distributional Reinforcement Learning for Satellite Inspection
November 2, 2026
A Quantile-Regression Deep Q-Network (QR-DQN) agent enables risk-sensitive multi-target inspection missions in low Earth orbit by modeling the full distribution of mission returns. The approach uses a mean-variance criterion to select inspection targets, accounting for the variability in low-thrust orbital transfers.
HOW THIS AFFECTS YOU
●
builderYou can implement distributional RL to handle safety-critical decision-making in environments with high outcome uncertainty.
●
researcherThis applies QR-DQN to complex, continuous orbital mechanics problems for more robust mission planning.