Publications Journal article · Delay-tolerant networks · 2024

Optimizing deep-space DTN congestion control via deep reinforcement learning

Lei Yang, Juan A. Fraire, Kanglian Zhao, Ruhai Wang, Wenfeng Li, and Hong Yang

Elsevier Computer Networks, vol. 255, p. 110865, 2024

Abstract

This paper introduces an innovative congestion control mechanism for delay/disruption-tolerant networking (DTN) within deep-space communication systems, leveraging the nuanced capabilities of deep reinforcement learning (DRL). This approach significantly departs from traditional methods, addressing the unique challenges of deep-space data transmissions. The proposed DRL-based strategy demonstrates a superior balance of critical factors, including transmission delay, energy efficiency, and data reception integrity. We assess our approach through meticulous simulation and comparison with established benchmark schemes. The findings underscore the mechanism's enhanced performance metrics, positing it as an appealing solution in the evolving landscape of non-terrestrial networking.

Cite

DOI 10.1016/j.comnet.2024.110865

BibTeX

@article{DBLP:journals/cn/YangFZWLY24,
  author  = {Lei Yang and Juan A. Fraire and Kanglian Zhao and Ruhai Wang and Wenfeng Li and Hong Yang},
  title   = {Optimizing deep-space {DTN} congestion control via deep reinforcement learning},
  journal = {Elsevier Computer Networks},
  volume  = {255},
  pages   = {110865},
  year    = {2024},
  doi     = {10.1016/j.comnet.2024.110865}
}

More on delay-tolerant networks

Autonomous Max-Flow Interplanetary Laser Link Scheduling for Martian Exploration Jason Gerard, Andre Ibrahim, Juan A. Fraire, and Sandra Céspedes IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE), 2024
Abstract
New communication technologies like delay and disruption-tolerant networking (DTN) and free-space laser communications are being developed to handle massive planetary telemetry data. However, existing infrastructure cannot meet the scalability and autonomy demands of future space missions, resulting in 95% of data left on the mission planet. Manual scheduling of deep-space laser transmissions by operators is inefficient and error-prone, limiting mission parallelization, while autonomous scheduling, though promising, may limit optimal network routes reducing the network capacity. We hypothesize that autonomously scheduling contact opportunities will enable acquisition, tracking, and pointing (ATP) for free-space lasers and increase the capacity of interplanetary backhaul networks. In this work, we propose the Laser Link Scheduler (LLS), which combines a novel max-flow network capacity model and a decision-making algorithm for contact selection. We define a series of deep-space communication scenarios where data is collected from orbiters around Mars and backhauled to Earth. LLS is evaluated in simulations using contact plans generated based on realistic conditions and orbital dynamics. Our results confirm that autonomous contact scheduling increases network capacity by 1.5-2x, reducing operational costs and enabling highly parallel space exploration missions.

HAL Publisher
Comparing Statistical, Analytical, and Learning-Based Routing Approaches for Delay-Tolerant Networks Pedro R. D'Argenio, Juan A. Fraire, Arnd Hartmanns, and Fernando D. Raverta ACM Transactions on Modeling and Computer Simulation, 2025
Abstract
In delay-tolerant networks (DTNs) with uncertain contact plans, the communication episodes and their reliabilities are known a priori . To maximise the end-to-end delivery probability, a bounded network-wide number of message copies are allowed. The resulting multi-copy routing optimization problem is naturally modelled as a Markov decision process with distributed information. In this paper, we provide an in-depth comparison of three solution approaches: statistical model checking with scheduler sampling, the analytical RUCoP algorithm based on probabilistic model checking, and an implementation of concurrent Q-learning. We use an extensive benchmark set comprising random networks, scalable binomial topologies, and realistic ring-road low Earth orbit satellite networks. We evaluate the obtained message delivery probabilities as well as the computational effort. Our results show that all three approaches are suitable tools for obtaining reliable routes in DTN, and expose a tradeoff between scalability and solution quality.

HAL Publisher
Ring Road Networks: Access for Anyone Marius Feldmann, Juan A. Fraire, Felix Walter, and Scott C. Burleigh IEEE Communications Magazine, 2022
Abstract
Several billion people currently lack reliable access to the Internet and thus to a tremendous source of knowledge. In this article, we describe a field-tested communication approach combining CubeSat platforms and delay-tolerant networking (DTN) solutions to provide asynchronous connectivity to populations and regions that are underserved by the Internet. The resulting class of networks is known as ring road networks (RRNs), a networking approach that is built on technology developed for the construction of a solar system Internet. The necessary self-sufficiency of DTN nodes enables network access to be deployed incrementally at low cost, supporting communities that cannot be profitably served by Internet satellite constellations. We present the RRN architecture and evaluate the expected performance by means of simulations. Based on the latter, we discuss µD3TN: a lightweight and open source DTN protocol stack for RRNs and other DTN classes. µD3TN has been flight-tested in ESA's OPS-SAT in low Earth orbit during December 2020 and May 2021. This work discusses the experiment results as we validated the RRN approach in concrete application use cases. The reported outcomes motivate a new application domain.

HAL Publisher

Delay-tolerant networks: the topic page All 84 papers on this topic