In a simulated two-tier 5G heterogeneous network, an improved deep Q-network controller achieved the best observed mean multi-KPI balance among four evaluated handover mechanisms, according to research published in PLOS One. The study evaluated mobility management within dense macro-small-cell deployments, where mobile user equipment frequently traverses cells characterized by differing coverage areas, transmit powers, and propagation environments. These conditions typically create elevated risks of radio-link failure, handover failure, ping-pong handover, and unnecessary handover events.
Deploying ultra-dense small cells in 5G heterogeneous networks improves network capacity, coverage, and spectral efficiency, but it simultaneously introduces substantial operational complexity to mobility management. Conventional handover configurations rely on fixed settings, applying a single-value handover margin and time-to-trigger. These uniform parameters fail to provide an effective trade-off across varying user equipment velocities and different handover directions, aggravating mobility failures and connection drops during rapid transit across heterogeneous cell boundaries.
To address these challenges, the researchers evaluated a mobility-state- and handover-direction-aware handover control parameter optimization framework. The system retains the standard 3GPP Event A3 handover scheme, the established state-based handover flow, and the conventional radio-link-monitoring framework. The evaluation tested four controllers under identical simulation conditions: a fixed Event A3 baseline, an analytical adaptive controller, a tabular Q-learning model, and an improved deep Q-network model.
The analytical adaptive controller specifically adjusted the handover margin and time-to-trigger based on the velocity of the user equipment and the direction of handover, distinguishing between macro-to-small, small-to-macro, and same-tier cell transitions. In contrast, the machine learning controllers utilized a multi-KPI reward formulation to learn discrete actions for handover margins and time-to-trigger settings across the operating environment.
Across all evaluated user velocities, the tabular Q-learning controller demonstrated improvements over the fixed baseline across principal mobility-robustness key performance indicators. The improved deep Q-network delivered the most favorable mean multi-KPI performance overall, aligning with recent deep reinforcement learning findings in mobile network optimization while operating entirely within standard handover and radio-link-monitoring frameworks.
The researchers conducted their evaluation within a repeatable two-tier macro-small-cell simulation employing 3D geometry with modeled user height and 3GPP TR 38.901 Urban Macro and Urban Micro propagation models. Because the policy-level comparisons were derived from three evaluation runs, the authors noted that reported uncertainty measures must be interpreted descriptively, with conclusions constrained strictly to the tested simulation parameters.
Subject of Research: Machine learning optimization of handover control parameters in 5G heterogeneous networks
Article Title: Mobility-state- and handover-direction-aware analytical and machine learning-based HCP optimization in 5G heterogeneous networks
Article References: Panhwer, F., Shaikh, M. M., Ansari, B., Ameer, I., Elmagzoub, M. A., & Shaikh, A. (2026). Mobility-state- and handover-direction-aware analytical and machine learning-based HCP optimization in 5G heterogeneous networks. PLOS One, 21(11), e0362310. https://doi.org/10.1371/journal.pone.0362310
Image Credits: AI Generated
DOI: 10.1371/journal.pone.0362310
Keywords: 5G networks, heterogeneous networks, handover control parameters, deep Q-network, reinforcement learning, mobility management, small cells, Event A3, radio-link failure, ping-pong handover, handover margin, time-to-trigger, Q-learning, telecommunications, wireless networks, 3GPP TR 38.901
Cite Scienmag News
Teresa Odom. (November 12, 2026). Deep Q-Network Optimization Improves Handover Performance in Simulated 5G Heterogeneous Networks. Scienmag. https://scienmag.com/deep-q-network-optimization-improves-handover-performance-in-simulated-5g-heterogeneous-networks/
Teresa Odom. "Deep Q-Network Optimization Improves Handover Performance in Simulated 5G Heterogeneous Networks." Scienmag, 12 November 2026, https://scienmag.com/deep-q-network-optimization-improves-handover-performance-in-simulated-5g-heterogeneous-networks/. Accessed 16 November 2026.
Teresa Odom. "Deep Q-Network Optimization Improves Handover Performance in Simulated 5G Heterogeneous Networks." Scienmag. November 12, 2026. https://scienmag.com/deep-q-network-optimization-improves-handover-performance-in-simulated-5g-heterogeneous-networks/








