Quadrotor adaptive control improves tracking despite sudden disturbances
Dual Gradient Consensus for Adaptive Learning to Transient Out-of-Distribution Disturbances in Quadrotor Control
Robotics
Summary
Quadrotors often struggle to maintain smooth flight when unexpected disturbances occur. The authors introduce a new method called Dual Gradient Consensus (DGC) that better adapts to sudden changes in the environment by carefully adjusting how it learns from recent sensor data. This approach avoids mixing up reliable information with brief, unusual errors, resulting in more accurate control of the quadrotor's flight path. Tests showed that DGC outperformed existing methods in keeping the drone on its intended course.
What this means in practice
- •For drone control engineers: Improve quadrotor flight stability in environments with sudden unpredictable disturbances by selectively adapting control updates.
- •For autonomous vehicle developers: Enhance adaptive control systems to maintain performance during short-term unexpected changes in operating conditions.
Authors
Mingtian Tan, Vasileios Tzoumas
Abstract
Quadrotor trajectory tracking under unmodeled disturbances often relies on online residual adaptation, but short out-domain residuals can corrupt the learned disturbance model when constant-step updates treat every residual gradient uniformly. We propose Dual Gradient Consensus (DGC), an adaptive-learning method for mixed in-domain/out-domain disturbances that retains the random Fourier feature model, offline BPTT policy, and least-squares adaptation of a prior domain-adaptive residual policy (DAPL) while modifying the online update step size. DGC compares each residual gradient with long- and short-term gradient memories, classifies the current adaptation regime, and uses a confidence map to suppress unreliable updates while recovering adaptation after persistent domain shifts. We evaluate DGC in numerical quadrotor trajectory tracking under four residuals that mix an in-domain component with short out-domain events. Compared with $L_1$-augmented MPC, RFF-based MPC, constant-step DAPL, and a frozen-parameter baseline, DGC achieves the lowest mean position RMSE in all four cases.