5G6G-NetworkSlicing-QoSAllocation

Queueing-theory- and 3GPP-standard-grounded multi-slice (eMBB/URLLC/mMTC) radio resource allocation dataset: 20 cell configurations, 800 traffic-load episodes, SLA-violation labels, for offline resource-allocation policy and QoS-violation prediction research.

About Dataset

5G6G-NetworkSlicing-QoSAllocation simulates the three canonical 3GPP-defined network slice types (3GPP TS 23.501's 5QI framework) -- eMBB (enhanced Mobile Broadband, throughput-oriented), URLLC (Ultra-Reliable Low-Latency Communication, ~10ms packet-delay-budget and 10^-5-scale reliability targets), and mMTC (massive Machine-Type Communications, high connection density, latency-tolerant) -- sharing a finite pool of radio resource blocks across 20 independently-sampled cell configurations. Session arrivals per slice follow a Poisson process; packet-level queueing delay emerges from genuine M/G/1-style queueing dynamics as a function of utilization, not an assigned number. Physical-layer throughput per resource block follows the Shannon-Hartley capacity formula with per-interval SINR generated from log-normal shadowing plus fast-fading. Resource allocation is computed via a proportional-fair-weighted scheduling heuristic (a standard non-RL reference policy from the literature). Each of 800 traffic-load episodes (low/medium/high load mixes) is labeled with per-slice SLA violation flags whenever queueing delay or packet error rate exceeds that slice's standard-derived target. This is a fully synthetic dataset -- no real operator, subscriber, or proprietary network data was used.

Purpose of Dataset

This Dataset Supports Research On Network-slice Resource Allocation Policies (Offline Rl, Imitation Learning, Supervised Qos-violation Prediction) That Generalize Across Different Cell/channel Conditions, Grounded In Real Queueing Theory And 3gpp Qos Standards Rather Than Arbitrary Synthetic Traffic. It Targets Researchers And Engineers Building Admission-control And Resource-scheduling Policies For Multi-tenant 5g/6g Ran Slicing, Particularly Where Early Prediction Of An Impending Sla Violation -- Before It Happens -- Is The Operationally Useful Capability. The Cell-grouped Design Supports Studying Whether A Resource-allocation Policy Trained Under Some Radio Conditions Generalizes To A New Cell It Has Never Operated In.