Skip to content

Path Tracking

Path tracking converts geometric references into feasible attitude and thrust commands. This chapter spans linear and nonlinear feedback structures used for hover and mission-level behavior.

Core Questions

  • How should positional and attitude loops be decoupled?
  • Which error-state definitions improve transient response?
  • How do saturation and actuator limits affect stability margins?

Algorithms

The One Thing That Separates Hovering From Tracking

Every controller in this chapter can hold a point. What distinguishes them on a moving reference is whether they are given its derivatives.

A pure feedback loop has to manufacture the reference's own acceleration out of tracking error, so it settles at whatever error does the job — error proportional to how hard the trajectory is. Feeding the reference acceleration forward removes that term from the error dynamics entirely. On the atlas figure-8 the difference is an order of magnitude for every controller here:

ControllerFeedback onlyWith feed-forward
LQR0.158 m0.017 m
Geometric SO(3)0.388 m0.023 m
MPC (preview)0.395 m0.032 m
Feedback linearisation0.234 m0.014 m

The symptom to recognise: tracking error that scales with reference speed, and an overshoot on the trajectory's fastest axis. That is not a gain that needs raising.

Prerequisites

  • State-space control basics
  • Hover linearization around trim
  • Attitude representation in SO(3)/Euler form

Released under the MIT License.