Is this project an undergraduate, graduate, or faculty project?

Graduate

Project Type

group

Campus

Daytona Beach

Authors' Class Standing

Edison Alberto Martinez Samaniego, Graduate student Natalie Alexander, Graduate student Kaelyn Weddle, Graduate student

Lead Presenter's Name

Edison Alberto Martinez Samaniego

Lead Presenter's College

DB College of Engineering

Faculty Mentor Name

Dr. Gamage Dumindu Samaraweera

Abstract

Learning Motion Primitive Selection and Environment Abstraction Advanced Air Mobility (AAM) is emerging as a transformative solution for short and medium range transportation; however, it introduces an operational model that differs significantly from conventional aviation. AAM vehicles are expected to operate closer to populated areas, with increased autonomy, in dense urban and suburban environments. These settings present constrained maneuvering conditions which highlights the importance of maintaining safe operation under degraded flight conditions. Abnormal conditions may endanger onboard passengers, people on the ground, and surrounding infrastructure, making rapid detection and mitigation essential to prevent loss of control. Recent research has explored frameworks that enhance operational safety by explicitly integrating flight envelope limits in the real-time trajectory replanning. Classical sampling-based planners (e.g., RRT/RRT*) typically construct trees using straight line connections, followed by post processing techniques such as smoothing the path with Dubins airplane motion primitives to ensure dynamic feasibility. On the other hand, using motion primitives directly during planning ensures feasibility but introduces significant computational cost. Computing a Dubins airplane motion primitive requires evaluating approximately 30 candidate trajectories and selecting the solution with the lowest cost. This project proposes a data-driven framework to improve planning efficiency while preserving trajectory quality. Supervised learning models, including Random Forests, Neural Networks, and Naive Bayes classifiers, will predict the most promising motion primitives between aircraft states, reducing the number of evaluations. Additionally, clustering techniques such as DBSCAN will be applied to spatial obstacle data to reduce collision-checking complexity in dense environments. Using datasets generated from an existing trajectory planner, the expected outcomes include reduced planning time, decreased number of motion primitive evaluations, and minimal deviation in trajectory cost compared to baseline methods. This work investigates how data mining methods can enable real-time, dynamically feasible trajectory generation for AAM operations under constrained flight environments.

Did this research project receive funding support (Spark, SURF, Research Abroad, Student Internal Grants, Collaborative, Climbing, or Ignite Grants) from the Office of Undergraduate Research?

No

Share

COinS
 

Learning Motion Primitive Selection and Environment Abstraction

Learning Motion Primitive Selection and Environment Abstraction Advanced Air Mobility (AAM) is emerging as a transformative solution for short and medium range transportation; however, it introduces an operational model that differs significantly from conventional aviation. AAM vehicles are expected to operate closer to populated areas, with increased autonomy, in dense urban and suburban environments. These settings present constrained maneuvering conditions which highlights the importance of maintaining safe operation under degraded flight conditions. Abnormal conditions may endanger onboard passengers, people on the ground, and surrounding infrastructure, making rapid detection and mitigation essential to prevent loss of control. Recent research has explored frameworks that enhance operational safety by explicitly integrating flight envelope limits in the real-time trajectory replanning. Classical sampling-based planners (e.g., RRT/RRT*) typically construct trees using straight line connections, followed by post processing techniques such as smoothing the path with Dubins airplane motion primitives to ensure dynamic feasibility. On the other hand, using motion primitives directly during planning ensures feasibility but introduces significant computational cost. Computing a Dubins airplane motion primitive requires evaluating approximately 30 candidate trajectories and selecting the solution with the lowest cost. This project proposes a data-driven framework to improve planning efficiency while preserving trajectory quality. Supervised learning models, including Random Forests, Neural Networks, and Naive Bayes classifiers, will predict the most promising motion primitives between aircraft states, reducing the number of evaluations. Additionally, clustering techniques such as DBSCAN will be applied to spatial obstacle data to reduce collision-checking complexity in dense environments. Using datasets generated from an existing trajectory planner, the expected outcomes include reduced planning time, decreased number of motion primitive evaluations, and minimal deviation in trajectory cost compared to baseline methods. This work investigates how data mining methods can enable real-time, dynamically feasible trajectory generation for AAM operations under constrained flight environments.

 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.