Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning
This thesis investigates energy-efficient load balancing in homogeneous multi-band cellular networks through the joint design of user association (UA) and transmit power allocation (PA). The original mixed-integer nonlinear formulation is decomposed into two coupled yet tractable subproblems: a UA s...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
AUC Knowledge Fountain
2026
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613433522814976 |
|---|---|
| access_status_str | Open Access |
| author | El Soukkary, Ahmed Shoukry |
| author_browse | El Soukkary, Ahmed Shoukry |
| author_facet | El Soukkary, Ahmed Shoukry |
| author_sort | El Soukkary, Ahmed Shoukry |
| collection | Thesis |
| description | This thesis investigates energy-efficient load balancing in homogeneous multi-band cellular networks through the joint design of user association (UA) and transmit power allocation (PA). The original mixed-integer nonlinear formulation is decomposed into two coupled yet tractable subproblems: a UA stage and a PA stage for high-frequency bands. For UA, a SINR-ratio-based heuristic is proposed to prioritize users that are most sensitive to suboptimal band assignments, and it is benchmarked against a Max- SINR baseline. For PA, the high-band power control problem is addressed using reinforcement learning, where a Proximal Policy Optimization (PPO) agent learns power levels and band-activation decisions under QoS constraints while accounting for dropped users and load balancing via Jain’s fairness index. The framework is evaluated using extensive simulations under both LOS and NLOS propagation, and is further extended to mobility scenarios to study time-varying channels and handover behavior. Results show that PPO-based power control is a primary driver of energy efficiency gains, while the proposed UA provides complementary improvements in fairness and connectivity, with trade-offs that depend on propagation conditions and reward weighting. Sensitivity analyses (including varying user density and the fraction of NLOS segments along trajectories) characterize robustness and reveal regimes where the joint UA+PA design improves energy efficiency while controlling dropped users, fairness degradation, and handover dynamics. |
| format | Thesis |
| id | oai:fount.aucegypt.edu:etds-3802 |
| institution | American University in Cairo (Egypt) |
| last_indexed | 2026-06-10T12:36:04.472Z |
| license_str | Not specified — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from AUC Knowledge Fountain — bepress |
| publishDate | 2026 |
| publishDateRange | 2026 |
| publishDateSort | 2026 |
| publisher | AUC Knowledge Fountain |
| publisherStr | AUC Knowledge Fountain |
| record_format | dspace |
| source_str | AUC Knowledge Fountain — bepress |
| spelling | oai:fount.aucegypt.edu:etds-3802 Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning El Soukkary, Ahmed Shoukry This thesis investigates energy-efficient load balancing in homogeneous multi-band cellular networks through the joint design of user association (UA) and transmit power allocation (PA). The original mixed-integer nonlinear formulation is decomposed into two coupled yet tractable subproblems: a UA stage and a PA stage for high-frequency bands. For UA, a SINR-ratio-based heuristic is proposed to prioritize users that are most sensitive to suboptimal band assignments, and it is benchmarked against a Max- SINR baseline. For PA, the high-band power control problem is addressed using reinforcement learning, where a Proximal Policy Optimization (PPO) agent learns power levels and band-activation decisions under QoS constraints while accounting for dropped users and load balancing via Jain’s fairness index. The framework is evaluated using extensive simulations under both LOS and NLOS propagation, and is further extended to mobility scenarios to study time-varying channels and handover behavior. Results show that PPO-based power control is a primary driver of energy efficiency gains, while the proposed UA provides complementary improvements in fairness and connectivity, with trade-offs that depend on propagation conditions and reward weighting. Sensitivity analyses (including varying user density and the fraction of NLOS segments along trajectories) characterize robustness and reveal regimes where the joint UA+PA design improves energy efficiency while controlling dropped users, fairness degradation, and handover dynamics. 2026-06-15T07:00:00Z thesis application/pdf https://fount.aucegypt.edu/etds/2743 https://fount.aucegypt.edu/context/etds/article/3802/viewcontent/ahmed_shoukry_elsoukkary_thesis.pdf Theses and Dissertations AUC Knowledge Fountain Energy efficiency Resource allocation Network management Machine learning Reinforcement learning Cellular network Multi-band Electrical and Electronics Systems and Communications |
| spellingShingle | Energy efficiency Resource allocation Network management Machine learning Reinforcement learning Cellular network Multi-band Electrical and Electronics Systems and Communications El Soukkary, Ahmed Shoukry Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning |
| title | Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning |
| title_full | Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning |
| title_fullStr | Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning |
| title_full_unstemmed | Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning |
| title_short | Energy Efficient Load Balancing in Multi-band Cellular Networks via Reinforcement Learning |
| title_sort | energy efficient load balancing in multi band cellular networks via reinforcement learning |
| topic | Energy efficiency Resource allocation Network management Machine learning Reinforcement learning Cellular network Multi-band Electrical and Electronics Systems and Communications |
| url | https://fount.aucegypt.edu/etds/2743 https://fount.aucegypt.edu/context/etds/article/3802/viewcontent/ahmed_shoukry_elsoukkary_thesis.pdf |
| work_keys_str_mv | AT elsoukkaryahmedshoukry energyefficientloadbalancinginmultibandcellularnetworksviareinforcementlearning |