Working toward agile flight for small UAVs

As a professor of Mechanical Engineering at Carnegie Mellon University, Sarah Bergbreiter loves collaborating with researchers in other fields, and she is especially excited to collaborate with experts in biology, neuroscience, and dynamics for her new project working to achieve agile flight for small robots. Looking forward to her collaboration, Bergbreiter says, “There’s so much interesting research that happens at the interfaces of these disciplines in biology and engineering.”

Bergbreiter and her collaborators were one of 24 teams recently presented with a Department of Defense (DoD) multidisciplinary university research initiative (MURI) award. Bergbreiter and her collaborators will receive $1.5 million per year for their 5-year project, which will be headed by the University of Washington, with another collaborator at MIT, and Bergbreiter at CMU.

In her own research, Bergbreiter focuses on small-scale robotics because she enjoys the challenge they pose. Working in small robotics, Bergbreiter has to design all of her own motors and sensors instead of buying them off the shelf, and she has to think harder about where to put the robots’ control and intelligence. Bergbreiter says she enjoys the different obstacles and ways of thinking. Laughingly, she also suggested she might have been drawn to her field because of her smaller stature.

If a gust hits a moth, certain neurons start firing and the moth knows it needs to move its body so it stays stable and upright.

Sarah Bergbreiter, Professor, Mechanical Engineering, Carnegie Mellon University

Small robots have a number of practical applications, including use in medical and surgical operations. Unmanned aerial vehicles (UAVs) also have important applications, mostly tied to delivery systems, construction, and civil infrastructure. With her current collaborative project to advance agile flight, Bergbreiter and her team are looking to develop UAVs that can fly around outside more efficiently and deal with disturbances such as gusts or breezes.

Insects and other flying animals have thousands of mechanosensors all over their bodies that they use to help them fly. Moths, for example, have hundreds of strain sensors on their wings, but they do not get data from all of their sensors like a traditional robot. Bergbreiter’s biology collaborators have found that moths have filters built into their neural circuits so that their neurons only fire when they receive a particular signal. According to Bergbreiter, “If a gust hits a moth, certain neurons start firing and the moth knows it needs to move its body so it stays stable and upright.”

Currently, putting hundreds of sensors on a robot would lead to an overwhelming amount of data. For her collaborative project, Bergbreiter will investigate how to integrate such a large number of sensors onto her small robots for what she calls “feed-forward control,” which references robots’ ability to respond to sensor data. Bergbreiter will experiment with pushing the data filtering to the mechanical design of her robots versus filtering sensor data computationally. Depending on where Bergbreiter puts the system’s intelligence, she will see tradeoffs in speed, power, and complexity.

Bergbreiter and her collaborators’ research will hopefully result in robots that can fly around outside just as well as they can inside, overcoming unexpected gusts or breezes that would otherwise cause them to flip over and crash. Bergbreiter’s project could potentially have huge impacts on UAV delivery systems as well as on engineers’ abilities to inspect things such as wind turbines and jet engines to keep them running efficiently.

Media contact:
Lisa Kulick, lkulick@andrew.cmu.edu