Satellite Internet Speeds Boosted by Tackling Data Delays in 6G Networks

Researchers are increasingly focused on dual connectivity (DC) as a means of improving throughput and reliability in modern wireless networks. Achilles Machumilane and Alberto Gotta, both from the Institute of Information Science and Technologies (ISTI), CNR, alongside their colleagues, present a novel mathematical framework to address a critical challenge within DC deployments, specifically concerning packet loss in Low Earth Orbit (LEO) satellite constellations. This work is significant because it moves beyond empirical modelling by providing analytical metrics for average end-to-end packet loss when employing techniques like packet duplication, switching, or network coding. The resulting insights will enable the derivation of optimal policies and facilitate comparison with machine learning approaches, ultimately optimising resource allocation and performance in complex, dynamically changing network topologies.

Optimising 6G packet delivery with dual connectivity and Markov chain modelling enhances network reliability and efficiency

Scientists have developed a mathematical framework to optimise packet delivery in future 6G networks integrating terrestrial and non-terrestrial components. The research focuses on mitigating these issues in real-time traffic where timely delivery is paramount.
This study provides a precise method for calculating average end-to-end packet loss, modelling the wireless channel as a Discrete Markov Chain. The framework specifically evaluates the effectiveness of combining packet duplication and packet switching, alongside the application of network coding techniques, within a dual connectivity setup.

By accurately quantifying the impact of these strategies, researchers aim to derive optimal policies for data transmission. These policies can then be benchmarked against empirical models generated through Machine Learning algorithms, offering a pathway towards adaptive and efficient network management.

The investigation centres on Non-Terrestrial Networks (NTNs), a key element of the evolving 3GPP standard designed to extend mobile network coverage via satellite technology. The highly dynamic nature of LEO constellations, with their constantly changing topology and short visibility windows, introduces unique challenges for maintaining stable connections.

Average snapshot durations, measured through emulation of constellations like Starlink, are in the order of minutes, necessitating frequent handovers between satellites. This constant change impacts routing tables and introduces fluctuations in channel latency, potentially causing out-of-sequence flows and subsequent packet reordering.

Packet reordering negatively affects protocols like TCP, QUIC, and RTP, leading to degraded performance, retransmissions, and glitches in streaming applications. The research demonstrates how network coding, combined with packet duplication and switching, can effectively mitigate packet dropouts caused by delay jitter in dual connectivity scenarios.

Dual connectivity involves a user device simultaneously connecting to two independent access nodes, one acting as a Master Node and the other as a Slave Node, enabling more robust and adaptable communication. The framework considers both transparent and regenerative satellite payloads, assessing their impact on overall network performance and reliability.

Modelling Packet Loss Rate with Markov Chains under Dual Connectivity presents a comprehensive analytical framework

A Discrete Markov Chain model underpinned the calculation of average end-to-end packet loss rates in this study, evaluating performance across different dual connectivity scenarios. The research investigated packet duplication, packet switching, and network coding techniques to mitigate packet reordering and congestion in networks with varying path delays, such as those found in terrestrial and non-terrestrial integrated systems.

Performance was assessed through detailed numerical analysis, considering three satellite elevation angle scenarios: 70◦-60◦, 60◦-45◦, and 70◦-45◦. The methodology systematically varied key parameters to explore trade-offs between the three techniques. Load balance (LB) ratios between two satellites, S1 and S2, were adjusted, alongside the balance between packet duplication and packet switching via a designated DT ratio.

The average redundancy factor (RF) and coding rate were also manipulated to quantify their impact on end-to-end packet loss rate (E2E-PLR). Tables I, II, and III present the evaluation results, showcasing E2E-PLR values achieved with different parameter combinations. Notably, network coding consistently demonstrated a lower E2E-PLR compared to packet duplication and packet switching.

However, this improvement came with increased encoding and decoding complexity, potentially impacting information rate and introducing delay. For instance, with links at 70◦-60◦, achieving an E2E-PLR of 0.005 with packet duplication and packet switching required either reliance on a single satellite or a redundancy factor of approximately 1.8, reducing the information rate. In contrast, network coding could achieve the same target with RF = 2.5 when N = 10, or RF = 2 when N = 20, demonstrating a more efficient use of resources.

Packet loss mitigation via Markov chains and de-jitter buffer analysis in dual connectivity offers improved video quality

Researchers detail a mathematical framework for calculating average end-to-end packet loss in wireless channels employing Discrete Markov Chains when combining packet duplication and packet switching, or utilising network coding in dual connectivity scenarios. This work provides metrics for deriving optimal policies, intended for comparison with empirical models learned through machine learning algorithms.

The analysis focuses on mitigating packet loss stemming from reordering issues inherent in dual connectivity, particularly within non-terrestrial networks. Investigations into real 3G measurements analysed the impact of de-jitter buffers in RTP-based live streaming with multi-connectivity. The de-jitter buffer forwards the initial received copy of each packet, identified by its serial number, but differing path latencies introduce a non-null probability of exceeding maximum tolerated jitter thresholds.

Even without reordering, jitter in packet pacing within VoIP applications significantly degrades the user’s quality of experience. The study specifically examines 3GPP-NTN mechanisms to reduce packet losses linked to reordering in dual connectivity non-terrestrial networks, evaluating packet switching, packet splitting, packet duplication, and network coding techniques.

Mathematical analysis demonstrates an efficient approach to utilising network coding and combining packet duplication with packet splitting to mitigate packet dropouts caused by delay jitter across the two dual connectivity connections. The research explores various dual connectivity architectures, including those with transparent and regenerative payloads on satellites.

Configurations involving a UE connected via a transparent NTN-based NG-RAN and a cellular NG-RAN, with two ground-based gNBs connected via an Xn interface, were considered. Another architecture examined involved two transparent NTN-based NG-RANs, potentially utilising GEO or LEO satellites, where differing round-trip times can introduce delay discrepancies.

Regenerative architectures with gNB-DU onboard satellites connected to a terrestrial gNB-CU via F1 interfaces were also investigated, alongside NTN-NTN configurations with gNBs onboard connected via an Xn interface. The introduction of the NTN segment into the NG-RAN can cause traffic to experience different delays, potentially leading to packet reordering and dropouts if delays exceed established thresholds.

Packet loss prediction for dual connectivity incorporating heterogeneous network delays is a challenging research problem

Researchers have developed a mathematical framework for calculating average end-to-end packet loss in dual connectivity (DC) scenarios, a key feature of 5G evolution intended to enhance throughput and reliability. This framework accounts for differing delays between communication paths, such as those found in terrestrial and non-terrestrial integrated networks, or in networks utilising low Earth orbit (LEO) satellites.

The model considers the combined use of packet duplication, packet switching, and network coding techniques for traffic scheduling, providing a means to assess their effectiveness in mitigating packet reordering and congestion. The presented metrics enable the derivation of optimal policies based on a detailed understanding of the wireless channel’s loss process, allowing for comparison with empirical models learned through machine learning algorithms.

Evaluations utilising the model, specifically for dual connectivity to LEO satellites, demonstrate the potential for reducing end-to-end packet loss, although this is achieved at the cost of information rate and increased complexity. The authors acknowledge that the analysis did not address the encoding and decoding delays associated with higher levels of redundancy, representing a limitation of the current work.

Future research could explore optimisation methods such as deep reinforcement learning to determine optimal packet duplication rates in dynamic network conditions. This work establishes a valuable tool for designing efficient and reliable communication protocols in complex network topologies. By providing a means to accurately predict packet loss, the framework facilitates the development of strategies that balance performance gains with resource constraints. The ability to compare model-based policies with machine learning approaches offers a pathway towards adaptive and robust network management in future 5G and beyond systems.