Learning Wing Motions in a Flapping-Wing Air Vehicle

Students will explore the use of machine-learning to enable insect-like flapping-wing micro air vehicles to learn new wing motion schedules (wing gaits) to compensate for damage to the wings.  A sample vehicle is pictured below.  Minor damage to the wings will not cause immediate crash, but it will severely disrupt the vehicle’s ability to reliably hold to desired flight plans.  Thus, in-service adaptation is potentially useful in increasing system reliability and longevity.

Person's hand holding a mechanism

Recent generations of the device have adopted a new wing actuator that introduces new challenges in applying existing machine-learning methods developed in previous generations.  In this work, students will explore possible modifications to older learning methods to account for changes made to wing actuators. 

Depending on student preparedness and interest, one or Students may attempt one or more of the following tasks:

  • Measuring wing performance and making improvement to mathematical models of wing force generation
  • Experimentation with alternative machine-learning methods to learn new wing gaits
  • Experimentation with high-level automated control of the vehicle

Student will be provided with:

  • Extensively commented and documented simulation code written in C 
  • Basic instruction on the vehicle’s operation and control
  • Basic instruction on one core machine learning model
  • As available, access to the physical vehicle or data sets collected from it

Any one of the listed task areas may result in publishable research in appropriate conferences or journals.


Headshot of John Christopher Gallagher

John Christopher Gallagher

Professor, CEAS - Electrical and Computer Engineeri

Rhodes Hall