Goal-Oriented Bundle Management and Flow Control for Deep-Space Communications via Reinforcement Learning
Reinforcement learning manages deep-space bundle relays to cut excess age and improve flow fairness.
The paper models relay scheduling and dropping for the Bundle Protocol in a contact-driven deep-space network that forwards lunar and Martian updates through a shared buffer toward Earth. Freshness is measured by excess Age of Information, and a goal-oriented loss applies source-dependent penalties. Control is formulated as a constrained semi-Markov decision process and solved with masked Lagrangian PPO, which only permits bundles that can finish inside the remaining contact window. DTNSim simulations report improved age-based utility and inter-flow fairness versus FIFO and other baselines.
- Relay control is cast as a constrained semi-Markov decision process.
- Masked Lagrangian PPO schedules, drops, and shapes flows.
- Excess Age of Information captures mission-weighted freshness.
- DTNSim reports better utility and fairness than FIFO.
From the original article
1 min read at arxiv.orgDeep-space communications operate under intermittent connectivity, long propagation delays, and relay storage constraints. In such networks, the Bundle Protocol enables store-carry-forward communication, requiring relay nodes to manage limited storage and scarce contact opportunities.
By Sajjad Baghaeea · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.