Trajectory generation and tracking for drone racing

  • Jonathan Jamieson

Student thesis: Doctoral Thesis

Abstract

In this thesis trajectory generation for quadrotors, a type of rotorcraft UAV (Unmanned Aerial Vehicle), is considered with two diverent methods. The first applies the Maximum Principle of optimal control to derive closed-form analytical functions that describe the translational states for two different cases of nonholonomic constraints. Parametric optimisation is used to find the trajectories. Reachable sets are found numerically and a simple obstacle avoidance method is demonstrated for both cases. The second motion planning method found trajectories with polynomial basis functions that are parametrised by an abstract function between zero and one. This virtual time domain trajectory satisfied conditions placed on the boundary derivatives and followed a sequenceof desired waypoints. A process for finding a mapping function that converts the virtual domain trajectory into one on the standard time domain is developed to minimise the trajectory time whilst ensuring the motion remained feasible by enforcing bounds on the thrust required from each rotor. An algorithm that uses additional waypoints where necessary to ensure the trajectory does not collide with the gates that define the course is developed. A method for minimising the accumulated angular acceleration of the heading angle whilst remaining within a desired tolerance of the velocity vector angle is also described. Trajectory tracking is considered by modifying an existing quadrotor tracking controller on the Special Euclidean group SE(3) to include a Linear Extended State Observer that estimates and counteracts translational disturbances. The modified controller is shown to reduce the position tracking error in the presence of square wave, sinusoidal and wind disturbances.
Date of Award1 Apr 2017
LanguageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsEPSRC (Engineering and Physical Sciences Research Council)
SupervisorMalcolm Macdonald (Supervisor) & Ivan Glesk (Supervisor)

Cite this