This project shows a combination for drones autonomous navigation in dynamic environments. The algorithm combine Social GAN SGAN for human trajectory prediction with Rapidly-exploring Random Tree Star RRT* for path planning. The objective is to efficiently and safe navigate in the area with human. Drones would avoid moving human and maintaining optimal flight path. During training SGAN model, we use both public datasets and dataset collected in the lab, which improve its adaptability in the lab. This experiment was tested through simulations and real-word experiment. SGAN provided a good prediction of human trajectories, which help drones to adjust their path. RRT* would replan the path when potential collisions are detected. The combined algorithm had a high success rate in collision avoidance(80% in real-word experiments). Additionally, incorporating lab collected data improved the accuracy and reduce the average displacement error (ADE) and final displacement error (FDE). This study also faces some limitation. In real-world, the use of high-precision OptiTrack motion capture is not avaliable. Besides, the experiment area is small and in real world the environments are more complex. Future work should focus on improving the algorithm in larger and more complex environment. In conclusion, this research demonstrates the potential of combining deep learning-based path prediction with real-time path planning. It provide a robust solution for drones to safely navigate in dynamic environment.
","abstract_html":"<p>This project shows a combination for drones autonomous navigation in dynamic environments. The algorithm combine Social GAN SGAN for human trajectory prediction with Rapidly-exploring Random Tree Star RRT* for path planning. The objective is to efficiently and safe navigate in the area with human. Drones would avoid moving human and maintaining optimal flight path. During training SGAN model, we use both public datasets and dataset collected in the lab, which improve its adaptability in the lab. This experiment was tested through simulations and real-word experiment. SGAN provided a good prediction of human trajectories, which help drones to adjust their path. RRT* would replan the path when potential collisions are detected. The combined algorithm had a high success rate in collision avoidance(80% in real-word experiments). Additionally, incorporating lab collected data improved the accuracy and reduce the average displacement error (ADE) and final displacement error (FDE). This study also faces some limitation. In real-world, the use of high-precision OptiTrack motion capture is not avaliable. Besides, the experiment area is small and in real world the environments are more complex. Future work should focus on improving the algorithm in larger and more complex environment. In conclusion, this research demonstrates the potential of combining deep learning-based path prediction with real-time path planning. It provide a robust solution for drones to safely navigate in dynamic environment.</p>","abstract_has_math":false,"creators":["Zhou, Beichen"],"institution":null,"degree_name":"Master of Science (MS)","degree_level":"Thesis","degree_discipline":"Electrical & Systems Engineering","degree_department":null,"school":null,"contributors":["Ioannis (Yiannis) Kantaros","ShiNung Ching, Shen Zeng"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-12-01T08:00:00Z","date_published":"2024-12-01T08:00:00Z","updated_at":"2026-07-24T06:13:23Z","subjects":["crazyflie","SGAN","RRT*","Control","Motion prediction","path planning","Engineering"],"languages":["English (en)"],"rights":["I have not registered my thesis with the U.S. Copyright Office, and do not intend to."],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["https://openscholarship.wustl.edu/eng_etds/1117"],"render_values":[{"text":"https://openscholarship.wustl.edu/eng_etds/1117","href":"https://openscholarship.wustl.edu/eng_etds/1117","code":true}]}]},"links":{"outbound_url":"https://doi.org/10.7936/kncd-yk95","outbound_label":"DOI","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Ioannis (Yiannis) Kantaros","ShiNung Ching, Shen Zeng"]},{"key":"dc:creator","label":"Author","values":["Zhou, Beichen"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2025-06-05T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical & Systems Engineering","McKelvey School of Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science (MS)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["crazyflie","SGAN","RRT*","Control","Motion prediction","path planning","Engineering"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["English (en)"]},{"key":"dc:rights","label":"Dc Rights","values":["I have not registered my thesis with the U.S. Copyright Office, and do not intend to."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://doi.org/10.7936/kncd-yk95","https://openscholarship.wustl.edu/eng_etds/1117"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This project shows a combination for drones autonomous navigation in dynamic environments. The algorithm combine Social GAN SGAN for human trajectory prediction with Rapidly-exploring Random Tree Star RRT* for path planning. The objective is to efficiently and safe navigate in the area with human. Drones would avoid moving human and maintaining optimal flight path. During training SGAN model, we use both public datasets and dataset collected in the lab, which improve its adaptability in the lab. This experiment was tested through simulations and real-word experiment. SGAN provided a good prediction of human trajectories, which help drones to adjust their path. RRT* would replan the path when potential collisions are detected. The combined algorithm had a high success rate in collision avoidance(80% in real-word experiments). Additionally, incorporating lab collected data improved the accuracy and reduce the average displacement error (ADE) and final displacement error (FDE). This study also faces some limitation. In real-world, the use of high-precision OptiTrack motion capture is not avaliable. Besides, the experiment area is small and in real world the environments are more complex. Future work should focus on improving the algorithm in larger and more complex environment. In conclusion, this research demonstrates the potential of combining deep learning-based path prediction with real-time path planning. It provide a robust solution for drones to safely navigate in dynamic environment.
"]},{"key":"dc:title","label":"Title","values":["Human-aware Motion Planning for Aerial Robots"]}]}],"canonical_facts":{"dc:contributor":["Ioannis (Yiannis) Kantaros","ShiNung Ching, Shen Zeng"],"dc:creator":["Zhou, Beichen"],"dc:date.available":["2025-06-05T07:00:00Z"],"dc:description.abstract":["This project shows a combination for drones autonomous navigation in dynamic environments. The algorithm combine Social GAN SGAN for human trajectory prediction with Rapidly-exploring Random Tree Star RRT* for path planning. The objective is to efficiently and safe navigate in the area with human. Drones would avoid moving human and maintaining optimal flight path. During training SGAN model, we use both public datasets and dataset collected in the lab, which improve its adaptability in the lab. This experiment was tested through simulations and real-word experiment. SGAN provided a good prediction of human trajectories, which help drones to adjust their path. RRT* would replan the path when potential collisions are detected. The combined algorithm had a high success rate in collision avoidance(80% in real-word experiments). Additionally, incorporating lab collected data improved the accuracy and reduce the average displacement error (ADE) and final displacement error (FDE). This study also faces some limitation. In real-world, the use of high-precision OptiTrack motion capture is not avaliable. Besides, the experiment area is small and in real world the environments are more complex. Future work should focus on improving the algorithm in larger and more complex environment. In conclusion, this research demonstrates the potential of combining deep learning-based path prediction with real-time path planning. It provide a robust solution for drones to safely navigate in dynamic environment.
"],"dc:identifier":["https://doi.org/10.7936/kncd-yk95","https://openscholarship.wustl.edu/eng_etds/1117"],"dc:language":["English (en)"],"dc:rights":["I have not registered my thesis with the U.S. Copyright Office, and do not intend to."],"dc:subject":["crazyflie","SGAN","RRT*","Control","Motion prediction","path planning","Engineering"],"dc:title":["Human-aware Motion Planning for Aerial Robots"],"thesis:degree_discipline":["Electrical & Systems Engineering","McKelvey School of Engineering"],"thesis:degree_level":["Thesis"],"thesis:degree_name":["Master of Science (MS)"]},"updated_at":"2026-07-24T06:13:23Z"}