超可靠低时延通信与大规模机器类通信为5G重要的两大应用场景，而随着移动通信技术的发展，6G将要求同时实现这两种场景，如工业自动化、自动驾驶等新兴场景既需要保障低延时通信，又需要满足多用户接入的需求。基于此，本文利用多队列联合控制与短码域跨层调度的方法，对低延时无线通信展开研究。 首先，本文针对多用户场景下并行队列非正交多址接入系统，考虑随机业务到达与时变信道，研究了基于队列和信道状态的概率性跨层调度。本文通过受约束的马氏决策过程与对受控队列的稳态分析，得到了最优延时功率折中的理论极限，并且给出了基于用户队列信息与信道信息的服务速率与解码顺序。考虑到受约束的马氏决策空间变量会随用户数量指数增长，本文提出了基于克罗内克积的自动化状态转移矩阵生成算法，用以实现跨层策略的高效求解。 其次，本文针对多用户场景下并行队列协同多点传输系统，引入大规模天线阵列以提升覆盖范围，研究了联合用户小区选择与用户功率分配的跨层调度。本文基于李雅普诺夫优化与受控排队理论，给出了基于队列信息与信道信息的低导频开销调度策略。其中，本文将协同多点传输问题解耦成为用户小区选择问题、用户功率分配问题，利用大规模天线阵列的信道硬化与信道渐进正交特征简化优化问题，且基于虚拟权重与二分最大权重匹配等方法给出了调度策略。 然后，本文针对多用户场景下并行队列短码域定长编码系统，考虑低延时通信中编码码长受限特性，研究了多用户短码域定长编码跨层调度。本文联合考虑多用户、受控队列与有限长编码，并利用李雅普诺夫优化理论建立了最大化队列长度加权速率和的优化问题。其中，本文构建虚拟功率队列，利用队列稳定性条件满足了长时平均功率约束。最后，本文基于子问题解耦、卡尔丹公式与最优性条件等，给出了基于队列信息与信道信息的短码域定长编码策略。 最后，本文针对单用户场景下串行队列短码域变长编码系统，兼顾码长、码率和延时之间的平衡，研究了短码域变长编码的理论极限及其跨层调度。本文通过拓展数据队列，加入串联的符号队列，以短码速率公式为纽带，建立串行队列服务量和到达量的映射。然后，本文利用非线性的受约束马氏决策对服务速率与编码码长进行设计，拓展跨层调度的控制维度，取得了显著的平均延时降低。此外，通过空频分集与功率自适应，本文给出了以概率1满足硬延时约束的变长编码策略。
The emerging scenarios of industrial automation and autonomous driving pose significant challenges to future mobile communication technologies. Ultra-reliable low-latency communication and massive machine-type communication are among the three major application scenarios of 5G. However, in future 6G networks, these emerging scenarios will require both low-latency communication and multi-user access. Therefore, this thesis investigates low-latency wireless communication based on joint control of multiple queues and cross-layer scheduling of finite blocklength. Firstly, this thesis addresses probabilistic cross-layer scheduling based on queue and channel states in a non-orthogonal multiple access system with parallel queues in multi-user scenarios, considering random traffic arrivals and time-varying channels. By using constrained Markov decision processes and steady-state analysis of controlled queues, the theoretical limits of the optimal delay-power trade-off are derived. Additionally, service rates and decoding order based on user queue states and channel states are provided. Noting that the variable space of constrained Markov decision processes grows exponentially with the number of users, an automated transition matrix generation algorithm based on Kronecker product is proposed to achieve efficient cross-layer policy solutions. Secondly, this thesis studies coordinated multiple point transmission systems with parallel queues in multi-user scenarios. By introducing large-scale antenna arrays to enhance coverage, joint user-cell selection and user power allocation cross-layer scheduling is studied. Based on Lyapunov optimization and controlled queuing theory, low overhead pilot scheduling policies based on queue states and channel states are presented. The coordinated multiple point transmission problem is decoupled into user-cell selection and user power allocation problems. Using the channel hardening and asymptotic orthogonal characteristics of large-scale antenna arrays, the optimization problem is simplified, and scheduling policies are provided based on virtual weights and maximum weight matching. Thirdly, this thesis investigates finite blocklength coding cross-layer scheduling for low-latency communication under multi-user scenarios, considering the finite blocklength characteristics. By jointly considering multiple users, controlled queues, and finite blocklength coding, and employing Lyapunov optimization theory, a problem of maximizing the sum of queue-length weighted rates is formulated. Virtual power queues are constructed to satisfy long-term average power constraints using queue stability conditions. Finally, subproblem decoupling, Cardan‘s formula, and optimality conditions are used to provide finite blocklength coding scheduling policies based on queue states and channel states. Lastly, this thesis explores the theoretical limits and cross-layer scheduling of variable-length coding with single user, balancing coding blocklength, coding rate, and delay. By extending packet queue with serially connected symbol queue and establishing a mapping of service rate and arrival rate using the finite blocklength coding rate, the study utilizes nonlinear constrained Markov decision processes to design service rate and coding blocklength. Moreover, the study extend the control dimension of cross-layer scheduling, resulting in significant average delay reduction. Additionally, through space-frequency diversity and power adaptation, a policy that satisfies hard delay constraints with probability 1 is also presented.
