Cross-layer scheduling and routing for ultra reliable and low latency communication
Date
2024
Authors
Ren, Xiangyu
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The success of recent communication and computing technologies has attracted growing interest from both academia and industry in developing advanced applications such as metaverse, digital twin, and intelligent transport systems, which are characterized by ultra-high reliability and low latency communication (URLLC) requirements. More importantly, they require guaranteed services and their performance will degrade quickly if any metric is not satisfied,
However, the existing routing and scheduling solutions may not fully support these applications. We observe one major reason is the lack of collaboration between each layer in the protocol stack and inefficient handling of network information.
To bridge the gap, we leverage a novel protocol architecture SET which enables flexible protocol assembling via control function decomposition. A new concept called protocol control agent (PCA) is introduced to enable in-network intelligence and enhanced adaptability. PCA leverages cross-layer network information and collaborations to support QoS requirements and improve resource efficiency. In addition, we consider the freshness of network information for designing new protocols based on SET. To this end, we develop a range of scheduling and routing solutions to support ultra-reliable and low-latency communications in different networks.
We start with the end-to-end delay guarantee problem in wired data networks. A novel distributed solution named delay-guaranteed scheduling and routing protocol (DSRP) is proposed to provide packet-level delay guarantees and differentiated services within an autonomous system. In DSRP, we adopt priority queues with fixed buffer sizes to guarantee per-hop delay bounds and serve traffic with different delay requirements, and leverage path diversity for multiplexing gains.
To address the congestion issue, we introduce a congestion-aware mechanism where neighboring nodes exchange queue information to reflect their congestion status. Given the rich routing decision space, a scheduling algorithm based on renewal optimization named DSROpt is proposed to maximize overall network utility.
Next, we investigate the scheduling problem for time-critical applications in a single-hop downlink wireless network. For time-critical applications, being late is as severe as being dropped. However, packets are normally left with a small delay budget at the last hop and are likely to exceed their delay requirement when there is a standing queue. To address this issue, we propose a delay-aware selective scheduling (DASS) algorithm that selects packets to serve. In DASS, we introduce a new output gain model based on delay laxity and packet priority for characterizing each scheduling decision and formulated a multi-objective optimization problem to determine the optimal scheduling policy. Due to the problem's complexity and uncertainty of the network environment, a deep reinforcement learning framework is proposed to find the Pareto optimal scheduling decisions that maximize network utility and guarantee per-packet delay requirement.
While many new applications operate in mobile networks, it is crucial to design QoS-guaranteed solutions with the consideration of more challenges such as mobility and dynamic channel conditions. Thus, we extend our research to the vehicular network. To address these issues, we propose a hybrid control framework leveraging global and local network information and taking advantage of the time-scale difference between the change rate of network typology and channel condition to perform control at different scopes.
Based on the framework, we first focus on the routing problem in vehicular networks. A QoS-guaranteed clustering and routing protocol (QCRP) is proposed. In QCRP, the global information, i.e., network topology is used for clustering and route planning to ensure per-cluster connectivity and routing path initialization while the local information, i.e., channel condition and vehicle location are used for re-routing. Next, to support efficient in-network communication under the constraints of limited network resources, a QoS-guaranteed medium access control (QMAC) protocol to perform resource allocation is required. Following the same framework, a centralized spectrum allocation combined with distributed power and error control solution is proposed.
Overall, in this thesis, we introduce a new perspective of supporting stringent QoS requirements and demonstrate the effectiveness of our solution in various network settings. Our work opens up new possibilities for research extensions and applications.
Description
Keywords
URLLC, Network Optimization, QoS guarantee, Resource allocation