{"id":{"repo_id":"vt","oai_identifier":"oai:vtechworks.lib.vt.edu:10919/137539"},"canonical_url":"https://search.dev.ndltd.org/etd/vt/oai:vtechworks.lib.vt.edu:10919/137539","repository":{"repo_id":"vt","name":"Virginia Tech","base_url":"https://vtechworks.lib.vt.edu/oai/request"},"display":{"title":"Contemporary Approaches to Radar Detection: Robust Statistics, Reinforcement Learning, and Reconfigurable Intelligent Surfaces","abstract":"Radar detection is one of the most fundamental problems in wireless signal processing and has been thoroughly studied for decades. As such, there is very well established theory for radar detection under a myriad of different system models; however, as the machine that is research continues to churn, new advancements are made every day-- compelling the reliable, classical techniques, which serve as pillars of our understanding of radar signal processing, to adapt or improve. To this end, we explore the use of robust statistics, reinforcement learning (RL), and reconfigurable intelligent surfaces (RIS) to improve upon existing radar detection techniques. This work focuses only on monostatic co-located MIMO radar and radar networks. First, we look at the role of robust detection statistics and RL in radar detection. Because RL is a type of online learning, RL based radar detection is well suited for dynamic environments; however, typically, constant false alarm rate (CFAR) radar detection either requires a priori knowledge of noise and clutter distributions or the collection of secondary data. We solve these issues by leveraging detection techniques developed for the massive MIMO (MMIMO) regime for CFAR detection in unknown noise and clutter distributions to perform CFAR detection in unknown noise and clutter distributions with fewer antennas. We provide necessary conditions under which our detection technique (statistic) is valid; furthermore, we pair this detection technique with RL to learn where targets are likely to appear in a predefined, fixed search area, to efficiently perform multi-target detection with MIMO radar. Secondly, we generalize such RL based multi-target detection techniques from a single co-located MIMO radar to a network of co-located MIMO radars which collaborate to efficiently perform detection in a given search area. We develop both centralized and decentralized signal fusion techniques for CFAR detection with radar networks in unknown noise and clutter distributions. We discuss the generalization of our RL algorithm for a single radar to a network of cognitive radars and provide a simulation study to analyze performance of centralized and decentralized detection. Finally, RIS is a new and exciting area of research in wireless communications and sensing. We leverage work showing improved performance in classical RIS-aided radar systems by introducing beyond diagonal RIS-aided radar. We develop beyond diagonal RIS (BD-RIS) configuration optimization specifically for a radar setting by leveraging prior art in BD-RIS optimization. Through simulation, we show improvements in performance over classical RIS-aided radar.","abstract_html":"Radar detection is one of the most fundamental problems in wireless signal processing and has been thoroughly studied for decades. As such, there is very well established theory for radar detection under a myriad of different system models; however, as the machine that is research continues to churn, new advancements are made every day-- compelling the reliable, classical techniques, which serve as pillars of our understanding of radar signal processing, to adapt or improve. To this end, we explore the use of robust statistics, reinforcement learning (RL), and reconfigurable intelligent surfaces (RIS) to improve upon existing radar detection techniques. This work focuses only on monostatic co-located MIMO radar and radar networks. First, we look at the role of robust detection statistics and RL in radar detection. Because RL is a type of online learning, RL based radar detection is well suited for dynamic environments; however, typically, constant false alarm rate (CFAR) radar detection either requires a priori knowledge of noise and clutter distributions or the collection of secondary data. We solve these issues by leveraging detection techniques developed for the massive MIMO (MMIMO) regime for CFAR detection in unknown noise and clutter distributions to perform CFAR detection in unknown noise and clutter distributions with fewer antennas. We provide necessary conditions under which our detection technique (statistic) is valid; furthermore, we pair this detection technique with RL to learn where targets are likely to appear in a predefined, fixed search area, to efficiently perform multi-target detection with MIMO radar. Secondly, we generalize such RL based multi-target detection techniques from a single co-located MIMO radar to a network of co-located MIMO radars which collaborate to efficiently perform detection in a given search area. We develop both centralized and decentralized signal fusion techniques for CFAR detection with radar networks in unknown noise and clutter distributions. We discuss the generalization of our RL algorithm for a single radar to a network of cognitive radars and provide a simulation study to analyze performance of centralized and decentralized detection. Finally, RIS is a new and exciting area of research in wireless communications and sensing. We leverage work showing improved performance in classical RIS-aided radar systems by introducing beyond diagonal RIS-aided radar. We develop beyond diagonal RIS (BD-RIS) configuration optimization specifically for a radar setting by leveraging prior art in BD-RIS optimization. Through simulation, we show improvements in performance over classical RIS-aided radar.","abstract_has_math":false,"creators":["Goradia, Nicholas Landon Kirit"],"institution":"Virginia Tech","degree_name":"Master of Science","degree_level":"masters","degree_discipline":"Electrical Engineering","degree_department":"Electrical Engineering","school":null,"contributors":[],"advisors":[],"committee_chairs":["Dhillon, Harpreet Singh","Buehrer, Richard M."],"committee_members":["Jakubisin, Daniel"],"year":2025,"date_issued":"2025-08-19","date_published":"2025-08-19","updated_at":"2026-07-22T22:19:15Z","subjects":["Cognitive Radar","Cognitive Radar Networks","MIMO Radar","Reinforcement Learning","Beyond Diagonal Reconfigurable Intelligent Surface","Radar Detection","Robust Detection"],"languages":["en"],"rights":["In Copyright"],"rights_urls":["http://rightsstatements.org/vocab/InC/1.0/"],"identifier_entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:44479"],"render_values":[{"text":"vt_gsexam:44479","href":null,"code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/10919/137539","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.committeechair","label":"Committee Chair","values":["Dhillon, Harpreet Singh","Buehrer, Richard M."]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Jakubisin, Daniel"]},{"key":"dc:contributor.department","label":"Department","values":["Electrical Engineering"]},{"key":"dc:creator","label":"Author","values":["Goradia, Nicholas Landon Kirit"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-08-20T08:01:23Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-08-20T08:01:23Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-08-19"]},{"key":"dc:publisher","label":"Institution","values":["Virginia Tech"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["masters"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Virginia Polytechnic Institute and State University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Cognitive Radar","Cognitive Radar Networks","MIMO Radar","Reinforcement Learning","Beyond Diagonal Reconfigurable Intelligent Surface","Radar Detection","Robust Detection"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["In Copyright"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://rightsstatements.org/vocab/InC/1.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:44479"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10919/137539"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Radar detection is one of the most fundamental problems in wireless signal processing and has been thoroughly studied for decades. As such, there is very well established theory for radar detection under a myriad of different system models; however, as the machine that is research continues to churn, new advancements are made every day-- compelling the reliable, classical techniques, which serve as pillars of our understanding of radar signal processing, to adapt or improve. To this end, we explore the use of robust statistics, reinforcement learning (RL), and reconfigurable intelligent surfaces (RIS) to improve upon existing radar detection techniques. This work focuses only on monostatic co-located MIMO radar and radar networks. First, we look at the role of robust detection statistics and RL in radar detection. Because RL is a type of online learning, RL based radar detection is well suited for dynamic environments; however, typically, constant false alarm rate (CFAR) radar detection either requires a priori knowledge of noise and clutter distributions or the collection of secondary data. We solve these issues by leveraging detection techniques developed for the massive MIMO (MMIMO) regime for CFAR detection in unknown noise and clutter distributions to perform CFAR detection in unknown noise and clutter distributions with fewer antennas. We provide necessary conditions under which our detection technique (statistic) is valid; furthermore, we pair this detection technique with RL to learn where targets are likely to appear in a predefined, fixed search area, to efficiently perform multi-target detection with MIMO radar. Secondly, we generalize such RL based multi-target detection techniques from a single co-located MIMO radar to a network of co-located MIMO radars which collaborate to efficiently perform detection in a given search area. We develop both centralized and decentralized signal fusion techniques for CFAR detection with radar networks in unknown noise and clutter distributions. We discuss the generalization of our RL algorithm for a single radar to a network of cognitive radars and provide a simulation study to analyze performance of centralized and decentralized detection. Finally, RIS is a new and exciting area of research in wireless communications and sensing. We leverage work showing improved performance in classical RIS-aided radar systems by introducing beyond diagonal RIS-aided radar. We develop beyond diagonal RIS (BD-RIS) configuration optimization specifically for a radar setting by leveraging prior art in BD-RIS optimization. Through simulation, we show improvements in performance over classical RIS-aided radar."]},{"key":"dc:description.abstractgeneral","label":"General Abstract","values":["Radar detection is one of the most fundamental problems in wireless signal processing and has been thoroughly studied for decades. As such, there is very well established theory for radar detection under a myriad of different system models; however, as the machine that is research continues to churn, new advancements are made every day-- compelling the reliable, classical techniques, which serve as pillars of our understanding of radar signal processing, to adapt or improve. Specifically, we leverage reinforcement learning-- a type of machine learning characterized by special computer algorithms which attempt to learn highly non-intuitive patterns in data in real time by exploiting mathematics-- to make improvements on the performance of radar systems for detecting targets. Secondly, we leverage the exciting research area of reconfigurable intelligent surfaces (RIS), special surfaces constructed specifically to direct signals which hit them into a desired direction, to work in tandem with radar for improved detection performance. In the type of radar we consider here, a signal is sent out into a search area. This signal reflects off of a target and travels back to the radar to be received; however, during this signal's journey, through the air, it can be affected by many different things. From the inherent electromagnetic radiation that emits from all things (generally described as noise) to bouncing off of leaves, or perhaps the ocean, these effects, or disturbances, are inherently random and change the original signal in unknown ways. Typically, using probability theory, assumptions are either made about these disturbances, or additional data is collected to estimate how these disturbances behave. This step is crucial because these disturbances can cause ``false alarms\" in detection if their behavior is not properly estimated. One could imagine that a reliable radar system aims to maximize the true detection performance against some number of false alarms. To this end, we leverage probability theory to carry detection by setting our own false alarm rate without knowing the disturbances. We do not make specific assumptions on the disturbances before hand, nor do we collect additional data to estimate the disturbances. We couple this detection method (robust statistic) with reinforcement learning to learn where targets are most likely to appear in a given search area. We also develop methods for a network or group of radars to work collaboratively and share their information to detect targets more efficiently. Finally, because detection performance relies primarily on the quality of the signal that is received, we leverage RIS to gain an additional path for the signal to travel to targets and back towards the radar. This additional path improves the quality of the received signal and therefore improves detection performance. Furthermore, we investigate different types of RIS where the elements of the surface are connected to all other elements of the surface, and show even further improvements with these \"fully connected\" RIS architectures."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Master of Science"]},{"key":"dc:format.medium","label":"Dc Format Medium","values":["ETD"]},{"key":"dc:title","label":"Title","values":["Contemporary Approaches to Radar Detection: Robust Statistics, Reinforcement Learning, and Reconfigurable Intelligent Surfaces"]}]}],"canonical_facts":{"dc:contributor.committeechair":["Dhillon, Harpreet Singh","Buehrer, Richard M."],"dc:contributor.committeemember":["Jakubisin, Daniel"],"dc:contributor.department":["Electrical Engineering"],"dc:creator":["Goradia, Nicholas Landon Kirit"],"dc:date.accessioned":["2025-08-20T08:01:23Z"],"dc:date.available":["2025-08-20T08:01:23Z"],"dc:date.issued":["2025-08-19"],"dc:description.abstract":["Radar detection is one of the most fundamental problems in wireless signal processing and has been thoroughly studied for decades. As such, there is very well established theory for radar detection under a myriad of different system models; however, as the machine that is research continues to churn, new advancements are made every day-- compelling the reliable, classical techniques, which serve as pillars of our understanding of radar signal processing, to adapt or improve. To this end, we explore the use of robust statistics, reinforcement learning (RL), and reconfigurable intelligent surfaces (RIS) to improve upon existing radar detection techniques. This work focuses only on monostatic co-located MIMO radar and radar networks. First, we look at the role of robust detection statistics and RL in radar detection. Because RL is a type of online learning, RL based radar detection is well suited for dynamic environments; however, typically, constant false alarm rate (CFAR) radar detection either requires a priori knowledge of noise and clutter distributions or the collection of secondary data. We solve these issues by leveraging detection techniques developed for the massive MIMO (MMIMO) regime for CFAR detection in unknown noise and clutter distributions to perform CFAR detection in unknown noise and clutter distributions with fewer antennas. We provide necessary conditions under which our detection technique (statistic) is valid; furthermore, we pair this detection technique with RL to learn where targets are likely to appear in a predefined, fixed search area, to efficiently perform multi-target detection with MIMO radar. Secondly, we generalize such RL based multi-target detection techniques from a single co-located MIMO radar to a network of co-located MIMO radars which collaborate to efficiently perform detection in a given search area. We develop both centralized and decentralized signal fusion techniques for CFAR detection with radar networks in unknown noise and clutter distributions. We discuss the generalization of our RL algorithm for a single radar to a network of cognitive radars and provide a simulation study to analyze performance of centralized and decentralized detection. Finally, RIS is a new and exciting area of research in wireless communications and sensing. We leverage work showing improved performance in classical RIS-aided radar systems by introducing beyond diagonal RIS-aided radar. We develop beyond diagonal RIS (BD-RIS) configuration optimization specifically for a radar setting by leveraging prior art in BD-RIS optimization. Through simulation, we show improvements in performance over classical RIS-aided radar."],"dc:description.abstractgeneral":["Radar detection is one of the most fundamental problems in wireless signal processing and has been thoroughly studied for decades. As such, there is very well established theory for radar detection under a myriad of different system models; however, as the machine that is research continues to churn, new advancements are made every day-- compelling the reliable, classical techniques, which serve as pillars of our understanding of radar signal processing, to adapt or improve. Specifically, we leverage reinforcement learning-- a type of machine learning characterized by special computer algorithms which attempt to learn highly non-intuitive patterns in data in real time by exploiting mathematics-- to make improvements on the performance of radar systems for detecting targets. Secondly, we leverage the exciting research area of reconfigurable intelligent surfaces (RIS), special surfaces constructed specifically to direct signals which hit them into a desired direction, to work in tandem with radar for improved detection performance. In the type of radar we consider here, a signal is sent out into a search area. This signal reflects off of a target and travels back to the radar to be received; however, during this signal's journey, through the air, it can be affected by many different things. From the inherent electromagnetic radiation that emits from all things (generally described as noise) to bouncing off of leaves, or perhaps the ocean, these effects, or disturbances, are inherently random and change the original signal in unknown ways. Typically, using probability theory, assumptions are either made about these disturbances, or additional data is collected to estimate how these disturbances behave. This step is crucial because these disturbances can cause ``false alarms\" in detection if their behavior is not properly estimated. One could imagine that a reliable radar system aims to maximize the true detection performance against some number of false alarms. To this end, we leverage probability theory to carry detection by setting our own false alarm rate without knowing the disturbances. We do not make specific assumptions on the disturbances before hand, nor do we collect additional data to estimate the disturbances. We couple this detection method (robust statistic) with reinforcement learning to learn where targets are most likely to appear in a given search area. We also develop methods for a network or group of radars to work collaboratively and share their information to detect targets more efficiently. Finally, because detection performance relies primarily on the quality of the signal that is received, we leverage RIS to gain an additional path for the signal to travel to targets and back towards the radar. This additional path improves the quality of the received signal and therefore improves detection performance. Furthermore, we investigate different types of RIS where the elements of the surface are connected to all other elements of the surface, and show even further improvements with these \"fully connected\" RIS architectures."],"dc:description.degree":["Master of Science"],"dc:format.medium":["ETD"],"dc:identifier.other":["vt_gsexam:44479"],"dc:identifier.uri":["https://hdl.handle.net/10919/137539"],"dc:language.iso":["en"],"dc:publisher":["Virginia Tech"],"dc:rights":["In Copyright"],"dc:rights.uri":["http://rightsstatements.org/vocab/InC/1.0/"],"dc:subject":["Cognitive Radar","Cognitive Radar Networks","MIMO Radar","Reinforcement Learning","Beyond Diagonal Reconfigurable Intelligent Surface","Radar Detection","Robust Detection"],"dc:title":["Contemporary Approaches to Radar Detection: Robust Statistics, Reinforcement Learning, and Reconfigurable Intelligent Surfaces"],"dc:type":["Thesis"],"thesis:degree_discipline":["Electrical Engineering"],"thesis:degree_level":["masters"],"thesis:degree_name":["Master of Science"],"thesis:institution_name":["Virginia Polytechnic Institute and State University"]},"updated_at":"2026-07-22T22:19:15Z"}