Extracting a Quadcopter's Dynamics from Simulated Flight Data

What is it about?

We used a method called SINDy to discover the equations describing how a quadcopter drone flies just from studying measurements from practice flights in simulations. We first made the drone fly different routes to collect data. SINDy analyzed terms like speed and height over time, picking the most important to match the data without knowing the real equations. Testing on new routes showed the found equations worked as well as the real ones. This shows we can determine complex rules for physical systems just from observation data, which helps when the rules are too hard to calculate from first steps.

Featured Image

Why is it important?

- It demonstrates a data-driven approach to model discovery that does not require first-principles mathematical derivation of equations. This is useful when systems are too complex to model analytically. - Being able to determine models solely from observational data opens up modeling of real-world physical phenomena that are mathematically intractable, like weather, biology, materials science, etc. - For applications like drone control and design, having an accurate model of the dynamics is crucial. This work shows we can obtain high-fidelity models directly from flight test data.