Low-cost adaptive obstacle avoidance trajectory control for express delivery drone
Yanhui Zhang, Caisheng Wei, Yifan Zhang, Congcong Tian, Weifang Chen
TL;DR
This work addresses obstacle avoidance trajectory control (OATC) for express-delivery quadcopters with uncertain payloads and dynamic conditions. It presents a nonlinear variable-gain PID (NLVG-PID) controller augmented by extremum seeking (ES) to learn optimal gains, with gains $f_{K_p}^j$, $f_{K_i}^j$, $f_{K_d}^j$ bounded by $k_{ ext{min}}$ and $k_{ ext{max}}$. The ES framework estimates the gradient of the cost $\,\mathcal{J}(\vec{K})=\frac{1}{t_f-t_0}\int e^2(t,\vec{K})\mathrm{d}t$ and updates $\vec{K}$ via $\vec{K}(t)=\vec{K}(t-1)-\alpha \nabla\mathcal{J}(\vec{K}(t-1))$, with random restarts to avoid local minima. Simulations on storm-type and Lissajous 3D trajectories demonstrate reduced overshoot and faster settling times, indicating a low-cost, portable solution for delivery drones across varying wheelbases.
Abstract
This paper studies quadcopters obstacle avoidance trajectory control (OATC) problem for express delivery. A new nonlinear adaptive learning controller that is low-cost and portable to different wheelbase sizes is proposed to adapt to large-angle maneuvers and load changes in UAV delivery missions. The controller consists of a nonlinear variable gain (NLVG) function and an extreme value search (ES) algorithm to reduce overshoot and settling time. Finally, simulations were conducted on a quadcopter to verify the effectiveness of the proposed control scheme under two typical collision-free trajectories.