Adaptive Gain Nonlinear Observer for External Wrench Estimation in Human-UAV Physical Interaction

A team of researchers has published a novel algorithm that could make collaborative drones significantly more responsive and affordable. The paper, "Adaptive Gain Nonlinear Observer for External Wrench Estimation in Human-UAV Physical Interaction," introduces an Adaptive Gain Nonlinear Observer (AGNO) designed to accurately estimate the forces and torques (collectively called a "wrench") exerted by a human on a drone during physical collaboration, such as guiding a payload. Crucially, it achieves this without relying on expensive and heavy dedicated force-torque sensors, instead using the drone's own dynamic model and flight data.

The core innovation is the observer's explicit handling of the drone's non-constant inertia matrix, which is vital for real-world scenarios where payloads shift or drones have asymmetric designs. In comprehensive simulations of a two-quadrotor team carrying a shared payload, the AGNO demonstrated superior performance compared to a standard Extended Kalman Filter (EKF), showing lower root mean square errors (RMSE), especially for torque estimation under nonlinear interaction conditions. This rigorous approach, backed by Lyapunov stability analysis, paves the way for drones that can be intuitively guided by touch, making cooperative tasks like assisted cargo transport safer and more feasible by reducing both system complexity and cost.