{"id":{"repo_id":"vt","oai_identifier":"oai:vtechworks.lib.vt.edu:10919/135030"},"canonical_url":"https://search.dev.ndltd.org/etd/vt/oai:vtechworks.lib.vt.edu:10919/135030","repository":{"repo_id":"vt","name":"Virginia Tech","base_url":"https://vtechworks.lib.vt.edu/oai/request"},"display":{"title":"A Neuro-Symbolic Reinforcement Learning Architecture: Integrating Perception, Reasoning, and Control","abstract":"In recent years, neuro-symbolic learning methods have demonstrated promise in tasks re- quiring a semantic understanding that can often be missed by traditional deep learning techniques. By integrating symbolic reasoning with deep learning, neuro-symbolic architec- tures aim to be both interpretable and flexible. This thesis aims to apply neuro-symbolic learning to the domain of reinforcement learning. First, a simulation environment for robotic manipulation tasks is presented. In this environment, an analysis of policy-gradient-based reinforcement learning algorithms is given. Then, by leveraging the performance of deep learning with the semantic reasoning and interpretability of symbolically defined program- ming, a novel neuro-symbolic learning method is proposed to generalize tasks and motion planning for robotics applications using natural language. This novel neuro-symbolic can be seen as an adaptation of the Neuro-Symbolic Concept Learner[1] developed by IBM Wat- son, in which images and natural language are first processed by convolutional and residual neural networks, respectively, and then parsed by a symbolically reasoned program. Where the architecture proposed in this paper differs is in its use of the Neuro-Symbolic Concept Learner for preprocessing of a given input task, to then inform a reinforcement learning agent of how to act in a given environment. Finally, the novel adaptation of the Neuro-Symbolic Concept Learner is introduced as a method of demonstrating generalizable behavior through symbolic preprocessing.","abstract_html":"In recent years, neuro-symbolic learning methods have demonstrated promise in tasks re- quiring a semantic understanding that can often be missed by traditional deep learning techniques. By integrating symbolic reasoning with deep learning, neuro-symbolic architec- tures aim to be both interpretable and flexible. This thesis aims to apply neuro-symbolic learning to the domain of reinforcement learning. First, a simulation environment for robotic manipulation tasks is presented. In this environment, an analysis of policy-gradient-based reinforcement learning algorithms is given. Then, by leveraging the performance of deep learning with the semantic reasoning and interpretability of symbolically defined program- ming, a novel neuro-symbolic learning method is proposed to generalize tasks and motion planning for robotics applications using natural language. This novel neuro-symbolic can be seen as an adaptation of the Neuro-Symbolic Concept Learner[1] developed by IBM Wat- son, in which images and natural language are first processed by convolutional and residual neural networks, respectively, and then parsed by a symbolically reasoned program. Where the architecture proposed in this paper differs is in its use of the Neuro-Symbolic Concept Learner for preprocessing of a given input task, to then inform a reinforcement learning agent of how to act in a given environment. Finally, the novel adaptation of the Neuro-Symbolic Concept Learner is introduced as a method of demonstrating generalizable behavior through symbolic preprocessing.","abstract_has_math":false,"creators":["Ellis, Hunter Wayne"],"institution":"Virginia Tech","degree_name":"Master of Science","degree_level":"masters","degree_discipline":"Computer Engineering","degree_department":"Electrical and Computer Engineering","school":null,"contributors":[],"advisors":[],"committee_chairs":["Doan, Thinh Thanh","Hsiao, Michael S."],"committee_members":["Williams, Ryan K."],"year":2025,"date_issued":"2025-06-03","date_published":"2025-06-03","updated_at":"2026-07-22T22:20:40Z","subjects":["Reinforcement Learning","Neuro-Symbolic","Concept Learner","Robotics"],"languages":["en"],"rights":["Creative Commons Attribution 4.0 International"],"rights_urls":["http://creativecommons.org/licenses/by/4.0/"],"identifier_entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:43971"],"render_values":[{"text":"vt_gsexam:43971","href":null,"code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/10919/135030","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.committeechair","label":"Committee Chair","values":["Doan, Thinh Thanh","Hsiao, Michael S."]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Williams, Ryan K."]},{"key":"dc:contributor.department","label":"Department","values":["Electrical and Computer Engineering"]},{"key":"dc:creator","label":"Author","values":["Ellis, Hunter Wayne"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-06-04T08:03:26Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-06-04T08:03:26Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-06-03"]},{"key":"dc:publisher","label":"Institution","values":["Virginia Tech"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer 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":["Reinforcement Learning","Neuro-Symbolic","Concept Learner","Robotics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Creative Commons Attribution 4.0 International"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://creativecommons.org/licenses/by/4.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:43971"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10919/135030"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["In recent years, neuro-symbolic learning methods have demonstrated promise in tasks re- quiring a semantic understanding that can often be missed by traditional deep learning techniques. By integrating symbolic reasoning with deep learning, neuro-symbolic architec- tures aim to be both interpretable and flexible. This thesis aims to apply neuro-symbolic learning to the domain of reinforcement learning. First, a simulation environment for robotic manipulation tasks is presented. In this environment, an analysis of policy-gradient-based reinforcement learning algorithms is given. Then, by leveraging the performance of deep learning with the semantic reasoning and interpretability of symbolically defined program- ming, a novel neuro-symbolic learning method is proposed to generalize tasks and motion planning for robotics applications using natural language. This novel neuro-symbolic can be seen as an adaptation of the Neuro-Symbolic Concept Learner[1] developed by IBM Wat- son, in which images and natural language are first processed by convolutional and residual neural networks, respectively, and then parsed by a symbolically reasoned program. Where the architecture proposed in this paper differs is in its use of the Neuro-Symbolic Concept Learner for preprocessing of a given input task, to then inform a reinforcement learning agent of how to act in a given environment. Finally, the novel adaptation of the Neuro-Symbolic Concept Learner is introduced as a method of demonstrating generalizable behavior through symbolic preprocessing."]},{"key":"dc:description.abstractgeneral","label":"General Abstract","values":["Robots are becoming more capable, but teaching them to perform complex tasks in changing environments remains a major challenge. Traditional learning systems, like deep learning, are powerful, but they are often seen as black-boxes. This project explores a new approach that combines the strengths of deep learning with symbolic reasoning, which allows robots to reason about their actions and goals in a more human-interpretable way. In this thesis, a simulated environment was built for training and testing a robotic arm on object manipula- tion tasks using. A reinforcement learning system was developed to allow the robot to learn through trial and error, improving its performance over time. To improve generalization and task understanding, a new hybrid model was proposed that combines deep learning with symbolic logic. Inspired by IBM's Neuro-Symbolic Concept Learner, this model uses visual and language inputs to guide the robot's behavior based on symbolic representations. Unlike the original Concept Learner, this version is adapted to help a reinforcement learning agent decide what to do in a specific situation based on these symbolic cues. This research shows that combining symbolic reasoning with modern learning techniques could make robots more flexible, explainable, and capable of handling a wider variety of real-world tasks."]},{"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":["A Neuro-Symbolic Reinforcement Learning Architecture: Integrating Perception, Reasoning, and Control"]}]}],"canonical_facts":{"dc:contributor.committeechair":["Doan, Thinh Thanh","Hsiao, Michael S."],"dc:contributor.committeemember":["Williams, Ryan K."],"dc:contributor.department":["Electrical and Computer Engineering"],"dc:creator":["Ellis, Hunter Wayne"],"dc:date.accessioned":["2025-06-04T08:03:26Z"],"dc:date.available":["2025-06-04T08:03:26Z"],"dc:date.issued":["2025-06-03"],"dc:description.abstract":["In recent years, neuro-symbolic learning methods have demonstrated promise in tasks re- quiring a semantic understanding that can often be missed by traditional deep learning techniques. By integrating symbolic reasoning with deep learning, neuro-symbolic architec- tures aim to be both interpretable and flexible. This thesis aims to apply neuro-symbolic learning to the domain of reinforcement learning. First, a simulation environment for robotic manipulation tasks is presented. In this environment, an analysis of policy-gradient-based reinforcement learning algorithms is given. Then, by leveraging the performance of deep learning with the semantic reasoning and interpretability of symbolically defined program- ming, a novel neuro-symbolic learning method is proposed to generalize tasks and motion planning for robotics applications using natural language. This novel neuro-symbolic can be seen as an adaptation of the Neuro-Symbolic Concept Learner[1] developed by IBM Wat- son, in which images and natural language are first processed by convolutional and residual neural networks, respectively, and then parsed by a symbolically reasoned program. Where the architecture proposed in this paper differs is in its use of the Neuro-Symbolic Concept Learner for preprocessing of a given input task, to then inform a reinforcement learning agent of how to act in a given environment. Finally, the novel adaptation of the Neuro-Symbolic Concept Learner is introduced as a method of demonstrating generalizable behavior through symbolic preprocessing."],"dc:description.abstractgeneral":["Robots are becoming more capable, but teaching them to perform complex tasks in changing environments remains a major challenge. Traditional learning systems, like deep learning, are powerful, but they are often seen as black-boxes. This project explores a new approach that combines the strengths of deep learning with symbolic reasoning, which allows robots to reason about their actions and goals in a more human-interpretable way. In this thesis, a simulated environment was built for training and testing a robotic arm on object manipula- tion tasks using. A reinforcement learning system was developed to allow the robot to learn through trial and error, improving its performance over time. To improve generalization and task understanding, a new hybrid model was proposed that combines deep learning with symbolic logic. Inspired by IBM's Neuro-Symbolic Concept Learner, this model uses visual and language inputs to guide the robot's behavior based on symbolic representations. Unlike the original Concept Learner, this version is adapted to help a reinforcement learning agent decide what to do in a specific situation based on these symbolic cues. This research shows that combining symbolic reasoning with modern learning techniques could make robots more flexible, explainable, and capable of handling a wider variety of real-world tasks."],"dc:description.degree":["Master of Science"],"dc:format.medium":["ETD"],"dc:identifier.other":["vt_gsexam:43971"],"dc:identifier.uri":["https://hdl.handle.net/10919/135030"],"dc:language.iso":["en"],"dc:publisher":["Virginia Tech"],"dc:rights":["Creative Commons Attribution 4.0 International"],"dc:rights.uri":["http://creativecommons.org/licenses/by/4.0/"],"dc:subject":["Reinforcement Learning","Neuro-Symbolic","Concept Learner","Robotics"],"dc:title":["A Neuro-Symbolic Reinforcement Learning Architecture: Integrating Perception, Reasoning, and Control"],"dc:type":["Thesis"],"thesis:degree_discipline":["Computer Engineering"],"thesis:degree_level":["masters"],"thesis:degree_name":["Master of Science"],"thesis:institution_name":["Virginia Polytechnic Institute and State University"]},"updated_at":"2026-07-22T22:20:40Z"}