<p>Highly reliable low-latency communication (HRLLC), a key paradigm in 6G networks, aims to meet stringent requirements of ultra-low latency and extreme reliability, making it ideal for enabling delay-critical services. However, HRLLC faces challenges due to traffic uncertainties—unpredictable fluctuations that degrade quality of service (QoS) in dynamic environments. Existing optimization methods often rely on simplified network assumptions (precise knowledge of traffic arrivals) or suffer from prohibitive computational complexity (e.g., conditional value at risk, CVaR). This paper proposes a novel robust optimization framework that integrates routing, resource provisioning, and admission control for delay-critical services under uncertain network conditions. While leveraging the Bernstein approximation to handle traffic uncertainty, the proposed framework integrates robust routing, resource provisioning, and admission control into a unified architecture for large-scale HRLLC scenarios. This holistic design enables scalable and delay-aware decisionmaking beyond the scope of conventional Bernstein-based methods. The framework decomposes the robust optimization problem into subproblems, deriving closed-form solutions for resource allocation and employing a quantized dynamic programming algorithm to achieve an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$[{\cal{O}}(\epsilon),{\cal{O}}(1/{\epsilon}^{N+E})]$</EquationSource> <EquationSource Format="MATHML"><math display="block"> <mo stretchy="false">[</mo> <mrow> <mrow> <mi mathvariant="script">O</mi> </mrow> </mrow> <mo stretchy="false">(</mo> <mi>ϵ</mi> <mo stretchy="false">)</mo> <mo>,</mo> <mrow> <mrow> <mi mathvariant="script">O</mi> </mrow> </mrow> <mo stretchy="false">(</mo> <mn>1</mn> <mrow> <mo>/</mo> </mrow> <msup> <mrow> <mi>ϵ</mi> </mrow> <mrow> <mi>N</mi> <mo>+</mo> <mi>E</mi> </mrow> </msup> <mo stretchy="false">)</mo> <mo stretchy="false">]</mo> </math></EquationSource> </InlineEquation>-trade-off between optimality loss and computational complexity. Comprehensive packet-level simulations validate the effectiveness of our proposed framework, which outperforms existing methods such as OSPF and greedy admission control with respect to reliability, resource utilization, and admission control. By utilizing Bernstein approximation for resource allocation, the framework achieves a 99.99% delay guarantee and improves the success rate of service routing and admission control by up to 33.3%, with a 25.0% enhancement over benchmark approaches in resource utilization, demonstrating the potential of robust optimization to maintain QoS in dynamic, mission-critical applications.</p>

Robust joint optimization framework for highly reliable low-latency communication services under traffic uncertainties

期刊论文
发表日期：

2026年1月4日

查看全文

Yilin Ren,
Xinchen Lyu,
Xiaofeng Tao,
Keda Chen

摘要

Highly reliable low-latency communication (HRLLC), a key paradigm in 6G networks, aims to meet stringent requirements of ultra-low latency and extreme reliability, making it ideal for enabling delay-critical services. However, HRLLC faces challenges due to traffic uncertainties—unpredictable fluctuations that degrade quality of service (QoS) in dynamic environments. Existing optimization methods often rely on simplified network assumptions (precise knowledge of traffic arrivals) or suffer from prohibitive computational complexity (e.g., conditional value at risk, CVaR). This paper proposes a novel robust optimization framework that integrates routing, resource provisioning, and admission control for delay-critical services under uncertain network conditions. While leveraging the Bernstein approximation to handle traffic uncertainty, the proposed framework integrates robust routing, resource provisioning, and admission control into a unified architecture for large-scale HRLLC scenarios. This holistic design enables scalable and delay-aware decisionmaking beyond the scope of conventional Bernstein-based methods. The framework decomposes the robust optimization problem into subproblems, deriving closed-form solutions for resource allocation and employing a quantized dynamic programming algorithm to achieve an $[{\cal{O}}(\epsilon),{\cal{O}}(1/{\epsilon}^{N+E})]$ $[O (ϵ), O (1 / ϵ^{N + E})]$ -trade-off between optimality loss and computational complexity. Comprehensive packet-level simulations validate the effectiveness of our proposed framework, which outperforms existing methods such as OSPF and greedy admission control with respect to reliability, resource utilization, and admission control. By utilizing Bernstein approximation for resource allocation, the framework achieves a 99.99% delay guarantee and improves the success rate of service routing and admission control by up to 33.3%, with a 25.0% enhancement over benchmark approaches in resource utilization, demonstrating the potential of robust optimization to maintain QoS in dynamic, mission-critical applications.