Hierarchical Computing Architecture for mURLLC in Cell-Free Systems: Load Computation, Demand Mapping, and Resource Allocation

With the rapid growth of time-sensitive services, existing infrastructures struggle to guarantee large-scale millisecond-level latency and high reliability. To meet these demands, massive ultra-reliable and low-latency communication (mURLLC) has been introduced as a core scenario in sixth-generation (6G) networks. To address the demands of mURLLC, we propose a hierarchical computing architecture for cell-free (CF) systems, comprising layers for load computation, demand mapping, and resource allocation. Specifically, in the first layer, we develop network load computation methods for CF systems based on the grant-free random access (GFRA) mechanism, where the load is dynamically inferred from subchannel (SC) occupancy patterns. In the second layer, we establish an analytical model that links access failure probability with quality-of-service (QoS) requirements, enabling accurate mapping of subchannel resource demands. In the third layer, based on finite blocklength transmission theory, we establish a joint delay optimization model and formulate the end-to-end (E2E) instantaneous delay minimization problem. To effectively address this problem, we further develop an adaptive blocklength-based resource allocation (ABRA) scheme that integrates with the hierarchical computing architecture for minimizing E2E delay. Numerical results have demonstrated that the proposed architecture can accurately capture network load variations, automatically adjust blocklengths according to users’ QoS requirements, and significantly reduce E2E delay.