Abstract:
As the potential of unmanned aerial vehicles rapidly increases, there is a growing interest
in rotary vehicles as well as fixed wing. The quadrotor is small agile rotary vehicle controlled
by variable speed prop rotors. With no need for a swash plate the vehicle is low
cost as well as dynamically simple.
In order to achieve autonomous flight, any potential control algorithm must include trajectory
generation and trajectory following. Trajectory generation can be done using direct
or indirect methods. Indirect methods provide an optimal solution but are hard to solve
for anything other than the simplest of cases. Direct methods in comparison are often
sub-optimal but can be applied to a wider range of problems. Trajectory optimization
is typically performed within the control space, however, by posing the problem in the
output space, the problem can be simplified. Differential flatness is a property of some
dynamical systems which allows dynamic inversion and hence, output space optimization.
Trajectory following can be achieved through any number of linear control techniques,
this is demonstrated whereby a single trajectory is followed using LQR, this scheme is
limited however, as the vehicle is unable to adapt to environmental changes. Model based
predictive control guarantees constraint satisfaction at every time step, this however is
time consuming and therefore, a combined controller is proposed benefiting from the
adaptable nature of MBPC and the robustness and simplicity of LQR control.
There are numerous direct methods for trajectory optimization both in the output and
control space. Taranenko’s direct method has a number of benefits over other techniques,
including the use of a virtual argument, which separates the optimal path and the speed
problem. This enables the algorithm to solve the optimal time problem, the optimal fuel
problem or a combination of the two, without a deviation from the optimal path.
In order to implement such a control scheme, the issues of feedback, communication and
control action computation, require consideration. This work discusses the issues with
instrumentation and communication encountered when developing the control system and
provides open loop test results.
This work also extends the proposed control schemes to consider the problem of multiple
vehicle flight rendezvous. Specifically the problem of rendezvous when there is no communication
link, limited visibility and no agreed rendezvous point. Using Taranenko’s
direct method multiple vehicle rendezvous is simulated.