ArduPilot's new adaptive control handles unknown uncertainties across vehicles

Autopilots for aerial, ground, and marine vehicles face a tough challenge: they often encounter unstructured uncertainties, meaning not only are the system parameters unknown, but so is the way those parameters affect the dynamics. This makes it impossible to rely on a fixed structural model. Researchers Peng Li, Di Liu, and Simone Baldi have proposed an adaptive design that sidesteps this problem entirely. By leveraging only basic Euler-Lagrange properties—which hold across most mechanical systems regardless of their specific structure—their controller can adapt without structural knowledge of the uncertainties. The method is concise, practical, and aimed at unifying control across different vehicle types.

Validation was done in the ArduPlane module of ArduPilot, the popular open-source autopilot platform, and the same architecture applies directly to ArduCopter, ArduRover, and ArduSub. The paper, published in IEEE Transactions on Control Systems Technology, demonstrates enhanced performance compared to state-of-the-art methods that also address unstructured and state-dependent uncertainties. This unified approach promises simpler, more robust autopilot development across platforms, potentially reducing tuning effort and improving reliability in real-world missions where model accuracy is limited.