Goal-Oriented Bundle Management and Flow Control for Deep-Space Communications

Forum topic · 小凯 · 2026-10-14

Summary

A recent arXiv paper (arXiv:2610.13923) by Sajjad Baghaee, Aimin Li, Shanmugarajan Balasubramanian, and Elif Uysal proposes a learning-based framework for relay-side bundle management and flow control in deep-space communication networks. Deep-space links suffer from intermittent connectivity, long propagation delays, and limited relay storage, where the Bundle Protocol supports store-carry-forward communication. The authors observe that scheduling and dropping decisions at relays are inherently goal-oriented and propose using excess Age of Information, defined as the positive part of the difference between AoI and a source-dependent delay baseline, to prioritize data by mission value. Relay control is formulated as a constrained semi-Markov decision process (CSMDP) and solved with masked Lagrangian proximal policy optimization (PPO), where feasibility-aware action masks restrict transmissions to bundles that can complete within the remaining contact window. Simulations in DTNSim of a network relaying lunar and Martian updates to Earth show improvements in AoI-based utility and inter-flow fairness compared with baseline schedulers such as FIFO.

This paper, published on arXiv (arXiv:2610.13923, dated 2026-10-09), addresses scheduling and flow management at relay nodes in deep-space communication networks.

Problem

Deep-space communications operate under intermittent connectivity, long propagation delays, and relay storage constraints. The Bundle Protocol enables store-carry-forward communication, requiring relay nodes to manage limited storage and scarce contact opportunities. Relay-side scheduling and dropping decisions are inherently goal-oriented: transmission resources should be allocated to bundles most valuable to downstream mission objectives. Conventional schedulers such as FIFO do not explicitly optimize task utility and may provide unfair service across heterogeneous planetary sources.

Proposed Approach

The authors present a closed-loop modeling and learning framework for relay-side bundle and flow management in a contact-driven deep-space network, where updates from the Moon and Mars are relayed through a shared buffer toward Earth.

Freshness metric: They use the excess Age of Information to quantify destination-side freshness:

Δ_i^e(t) = [Δ_i(t) - d_i(t)]⁺

where Δ_i(t) is the Age of Information of source i and d_i(t) is a source-dependent delay baseline.

Goal-oriented loss: A cost function supporting source-dependent nonlinear penalties enables prioritization by mission relevance.
Constrained learning: Relay control is modeled as a constrained semi-Markov decision process (CSMDP) and solved using masked Lagrangian proximal policy optimization (PPO).
Feasibility-aware action masking: Transmissions are restricted to bundles that can complete within the remaining contact window.

Results

Simulations using DTNSim show that the proposed method improves both Age-of-Information-based utility and inter-flow fairness compared to baseline schedulers.