[Paper Review] Proactive Resilient Transmission and Scheduling Mechanisms for mmWave Networks

Mine Gokce Dogan, Martina Cardone+1 more|arXiv (Cornell University)|Nov 17, 2022

Millimeter-Wave Propagation and Modeling4 citations

TL;DR

This paper proposes proactive resilient transmission and scheduling mechanisms for mmWave networks that pre-configure multiple high-throughput paths to withstand link blockages. By combining a polynomial-time heuristic path selection with a Soft Actor-Critic DRL-based online adaptation, the framework achieves high end-to-end packet rates under dynamic blockages, outperforming alternatives in diverse topologies and channel conditions.

ABSTRACT

This paper aims to develop resilient transmission mechanisms to suitably distribute traffic across multiple paths in an arbitrary millimeter-wave (mmWave) network. The main contributions include: (a) the development of proactive transmission mechanisms that build resilience against network disruptions in advance, while achieving a high end-to-end packet rate; (b) the design of a heuristic path selection algorithm that efficiently selects (in polynomial time in the network size) multiple proactively resilient paths with high packet rates; and (c) the development of a hybrid scheduling algorithm that combines the proposed path selection algorithm with a deep reinforcement learning (DRL) based online approach for decentralized adaptation to blocked links and failed paths. To achieve resilience to link failures, a state-of-the-art Soft Actor-Critic DRL algorithm, which adapts the information flow through the network, is investigated. The proposed scheduling algorithm robustly adapts to link failures over different topologies, channel and blockage realizations while offering a superior performance to alternative algorithms.

Motivation & Objective

• Address the challenge of mmWave network resilience against frequent, unpredictable link blockages due to environmental obstructions.
• Develop proactive transmission mechanisms that pre-configure multiple resilient paths using only prior knowledge of link blockage probabilities.
• Design a scalable, polynomial-time path selection algorithm that identifies high-throughput, resilient paths in arbitrary mmWave networks.
• Integrate a deep reinforcement learning (DRL) component for decentralized, real-time adaptation to unexpected link failures and changing network conditions.
• Achieve optimal or near-optimal end-to-end packet rates under both worst-case and average-case blockage scenarios.

Proposed method

• Formulate optimal worst-case and average-case end-to-end packet rates using Linear Programs (LPs), with constraints on flow conservation and path activation.
• Propose a heuristic path selection algorithm based on a modified Dijkstra’s algorithm that prioritizes paths with high expected throughput under blockage probabilities.
• Introduce a hybrid scheduling framework combining proactive path selection with online DRL adaptation using the Soft Actor-Critic (SAC) algorithm.
• Use link-specific blockage probabilities and capacity values as inputs to compute path weights, enabling prioritization of robust, high-capacity paths.
• Implement a path pruning mechanism that removes the weakest link from the selected path after each iteration to diversify path selection and avoid congestion.
• Leverage the SAC DRL agent to dynamically reroute traffic in real time when links fail, without requiring full topology or channel state knowledge.

Experimental results

Research questions

• RQ1 How can we pre-configure multiple resilient paths in an mmWave network to maximize end-to-end packet rate under known link blockage probabilities?
• RQ2 What is the structural relationship between path diversity, link capacity, and resilience in mmWave networks with asymmetric blockage probabilities?
• RQ3 Can a heuristic path selection algorithm achieve near-optimal performance in polynomial time while ensuring resilience to multiple link failures?
• RQ4 How does a hybrid approach combining proactive path selection with online DRL adaptation outperform purely reactive or purely proactive schemes?
• RQ5 Under what conditions do overlapping paths outperform edge-disjoint paths in achieving optimal worst-case or average-case packet rates?

Key findings

• The optimal worst-case and average packet rates can be characterized via Linear Programs (LPs), with the worst-case solution being polynomial in network size but exponential in the number of blocked links.
• In 1-2-1 mmWave networks, only at most $2N+2$ paths are needed to achieve the optimal average packet rate, where $N$ is the number of relay nodes.
• When link capacities are equal, the optimal worst-case and average packet rates are always achieved by activating edge-disjoint paths.
• With unequal link capacities, overlapping paths can achieve better performance than edge-disjoint paths, and may reduce the variance of the achieved packet rate.
• The proposed heuristic path selection algorithm runs in polynomial time and achieves a feasible solution that satisfies all LP constraints, with provable optimality guarantees under equal capacity assumptions.
• The hybrid scheduling algorithm, combining proactive path selection with SAC-based online adaptation, robustly maintains high performance across diverse topologies and blockage realizations, outperforming alternative approaches in both stability and throughput.