{"id":{"repo_id":"rice","oai_identifier":"oai:repository.rice.edu:1911/118643"},"canonical_url":"https://search.dev.ndltd.org/etd/rice/oai:repository.rice.edu:1911/118643","repository":{"repo_id":"rice","name":"Rice University","base_url":"https://repository.rice.edu/server/oai/request"},"display":{"title":"Structure-utilized, Adaptive, and Efficient ML-based Proportional-Fair Scheduling in MIMO Networks for Non-stationary Channels","abstract":"Proportional Fair (PF) scheduling is widely used in multi-user MIMO systems to balance throughput and fairness. However, PF scheduling is an NP-hard problem, and hence, practical deployments approximate the optimal solution for lower latency at the cost of sub-optimal performance. More recently, machine learning (ML)-based approaches have demonstrated strong performance with low latency. However, ML-based methods typically assume stationary channel distributions, making them vulnerable to performance degradation under dynamic network conditions such as user mobility and location changes. In this work, I develop a new ML-based scheduling framework that adapts to non-stationary wireless conditions in real time. The framework adopts a Graph Neural Network (GNN)-based scheduler—which captures both user-specific metrics and inter-user interference patterns—enabling structurally sample-efficient learning that generalizes well across users and topologies. Complementing this, an adaptive control module called On-Demand and Online Learning (ODOL) detects distribution shifts and triggers fine-tuning using expert demonstrations. To further reduce adaptation latency, we introduce an efficient online data collection strategy guided by user mobility structure, which accelerates sample acquisition during online fine-tuning. Extensive evaluations using simulations and real-world channel traces demonstrate that the proposed method consistently maintains high spectral efficiency and fairness with rapid policy adaptation under evolving channel conditions, making it a practical solution for next-generation wireless networks.","abstract_html":"Proportional Fair (PF) scheduling is widely used in multi-user MIMO systems to balance throughput and fairness. However, PF scheduling is an NP-hard problem, and hence, practical deployments approximate the optimal solution for lower latency at the cost of sub-optimal performance. More recently, machine learning (ML)-based approaches have demonstrated strong performance with low latency. However, ML-based methods typically assume stationary channel distributions, making them vulnerable to performance degradation under dynamic network conditions such as user mobility and location changes. In this work, I develop a new ML-based scheduling framework that adapts to non-stationary wireless conditions in real time. The framework adopts a Graph Neural Network (GNN)-based scheduler—which captures both user-specific metrics and inter-user interference patterns—enabling structurally sample-efficient learning that generalizes well across users and topologies. Complementing this, an adaptive control module called On-Demand and Online Learning (ODOL) detects distribution shifts and triggers fine-tuning using expert demonstrations. To further reduce adaptation latency, we introduce an efficient online data collection strategy guided by user mobility structure, which accelerates sample acquisition during online fine-tuning. Extensive evaluations using simulations and real-world channel traces demonstrate that the proposed method consistently maintains high spectral efficiency and fairness with rapid policy adaptation under evolving channel conditions, making it a practical solution for next-generation wireless networks.","abstract_has_math":false,"creators":["Cheng, Yirong"],"institution":"Rice University","degree_name":"Doctor of Philosophy","degree_level":"Doctoral","degree_discipline":"Engineering","degree_department":null,"school":null,"contributors":[],"advisors":["Sabharwal, Ashutosh"],"committee_chairs":[],"committee_members":["Segarra, Santiago","Ng, Eugene T.S."],"year":2025,"date_issued":"2025-07-22","date_published":"2025-07-22","updated_at":"2026-07-24T04:10:34Z","subjects":["Behavior Cloning","Channel Non-stationarity","Change Detection","Graph Neural Network","Machine Learning","MIMO Scheduling","Proportional Fairness","Wireless Networks"],"languages":["eng"],"rights":["Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1911/118643","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Sabharwal, Ashutosh"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Segarra, Santiago","Ng, Eugene T.S."]},{"key":"dc:creator","label":"Author","values":["Cheng, Yirong"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-09-03T21:32:43Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-07-22"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Doctoral"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctor of Philosophy"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Rice University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Behavior Cloning","Channel Non-stationarity","Change Detection","Graph Neural Network","Machine Learning","MIMO Scheduling","Proportional Fairness","Wireless Networks"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1911/118643"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Proportional Fair (PF) scheduling is widely used in multi-user MIMO systems to balance throughput and fairness. However, PF scheduling is an NP-hard problem, and hence, practical deployments approximate the optimal solution for lower latency at the cost of sub-optimal performance. More recently, machine learning (ML)-based approaches have demonstrated strong performance with low latency. However, ML-based methods typically assume stationary channel distributions, making them vulnerable to performance degradation under dynamic network conditions such as user mobility and location changes. In this work, I develop a new ML-based scheduling framework that adapts to non-stationary wireless conditions in real time. The framework adopts a Graph Neural Network (GNN)-based scheduler—which captures both user-specific metrics and inter-user interference patterns—enabling structurally sample-efficient learning that generalizes well across users and topologies. Complementing this, an adaptive control module called On-Demand and Online Learning (ODOL) detects distribution shifts and triggers fine-tuning using expert demonstrations. To further reduce adaptation latency, we introduce an efficient online data collection strategy guided by user mobility structure, which accelerates sample acquisition during online fine-tuning. Extensive evaluations using simulations and real-world channel traces demonstrate that the proposed method consistently maintains high spectral efficiency and fairness with rapid policy adaptation under evolving channel conditions, making it a practical solution for next-generation wireless networks."]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Structure-utilized, Adaptive, and Efficient ML-based Proportional-Fair Scheduling in MIMO Networks for Non-stationary Channels"]}]}],"canonical_facts":{"dc:contributor.advisor":["Sabharwal, Ashutosh"],"dc:contributor.committeemember":["Segarra, Santiago","Ng, Eugene T.S."],"dc:creator":["Cheng, Yirong"],"dc:date.accessioned":["2025-09-03T21:32:43Z"],"dc:date.issued":["2025-07-22"],"dc:description.abstract":["Proportional Fair (PF) scheduling is widely used in multi-user MIMO systems to balance throughput and fairness. However, PF scheduling is an NP-hard problem, and hence, practical deployments approximate the optimal solution for lower latency at the cost of sub-optimal performance. More recently, machine learning (ML)-based approaches have demonstrated strong performance with low latency. However, ML-based methods typically assume stationary channel distributions, making them vulnerable to performance degradation under dynamic network conditions such as user mobility and location changes. In this work, I develop a new ML-based scheduling framework that adapts to non-stationary wireless conditions in real time. The framework adopts a Graph Neural Network (GNN)-based scheduler—which captures both user-specific metrics and inter-user interference patterns—enabling structurally sample-efficient learning that generalizes well across users and topologies. Complementing this, an adaptive control module called On-Demand and Online Learning (ODOL) detects distribution shifts and triggers fine-tuning using expert demonstrations. To further reduce adaptation latency, we introduce an efficient online data collection strategy guided by user mobility structure, which accelerates sample acquisition during online fine-tuning. Extensive evaluations using simulations and real-world channel traces demonstrate that the proposed method consistently maintains high spectral efficiency and fairness with rapid policy adaptation under evolving channel conditions, making it a practical solution for next-generation wireless networks."],"dc:format.mimetype":["application/pdf"],"dc:identifier.uri":["https://hdl.handle.net/1911/118643"],"dc:language.iso":["eng"],"dc:rights":["Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder."],"dc:subject":["Behavior Cloning","Channel Non-stationarity","Change Detection","Graph Neural Network","Machine Learning","MIMO Scheduling","Proportional Fairness","Wireless Networks"],"dc:title":["Structure-utilized, Adaptive, and Efficient ML-based Proportional-Fair Scheduling in MIMO Networks for Non-stationary Channels"],"dc:type":["Thesis"],"thesis:degree_discipline":["Engineering"],"thesis:degree_level":["Doctoral"],"thesis:degree_name":["Doctor of Philosophy"],"thesis:institution_name":["Rice University"]},"updated_at":"2026-07-24T04:10:34Z"}