TRACE: Trajectory Refinement with Control Error Enables Safe and Accurate Maneuvers

Zihan Yang, Jindou Jia, Yuhang Liu, Kexin Guo, Xiang Yu, Lei Guo

Abstract:

Normal trajectory optimizations are carried out
without the consideration of the closed-loop feasibility
and performance.
During aggressive flights, bias from the nominal trajectory
could trigger feedback with
larger control inputs,
which might break through the actuator constraints,
inducing stability issues.
By leveraging the idea of
TRAjectory refinement with Control Error (TRACE)
for higher closed-loop performance,
we develop a post-trajectory optimization approach
for safe and accurate tracking control.
A closed-loop model is established
combining both control policy and saturations in order
to model the dynamic progress of trajectory tracking.
Subsequently, a time-varying optimal control problem (OCP)
is constructed to enhance the tracking performance
by weighted minimizing the tracking error and the variation
from the nominal trajectory progress.
Meanwhile, a flatness-based method is presented to provide
aggressive time-optimal trajectories with
high-order derivatives of the reference states.
Examples show the effectiveness of our approach
in refining aggressive trajectories, where safe and accurate tracking
can be done using a baseline controller without fine-tuning and
further control allocation design.