Incremental Control of Hybrid Micro Air Vehicles

Abstract

Micro Air Vehicles (MAVs) can perform many useful tasks, such as mapping and delivery. For these tasks either rotorcraft are used, which can hover but are not very efficient, or fixed wing vehicles, which are efficient but can not hover. Hybrid MAVs combine the hovering of a rotorcraft with the efficiency of a fixed wing. The reason that these vehicles are not yet widely adopted is that they are very difficult to control.

This thesis addresses the use of Incremental Nonlinear Dynamic Inversion (INDI) for the control of the attitude and velocity of hybrid MAVs. This control method had not been applied in a real world application prior to this thesis, which is why the thesis encompasses the application to a quadrotor at first, which is easier to control than a hybrid MAV.

First, an INDI structure is proposed for the control of the angular accelerations of a quadrotor. I show that the delay that filtering of the angular acceleration produces should also be applied to the measurement of the actuator state. If this is done, the filtering does not appear in the transfer function from virtual control to angular acceleration, which turns out to be equal to the actuator dynamics. It is also shown that a disturbance, or unmodeled dynamics, is compensated with the transfer function of the actuator dynamics multiplied with the applied filter and a unit delay. Finally, it is shown how the effects of propeller inertia, which can be very significant in the yaw axis, can be dealt with and how the control effectiveness can be made adaptive. All these findings are validated with experiments on a Bebop quadrotor.

Second, this thesis includes a Weighted Least Squares (WLS) control allocation algorithm with priority management into the INDI controller. This means that for vehicles with coupled control effectors, certain control objectives can be given pri- ority upon actuator saturation. This is very important for vehicles with controlled axes that are not very important for the stability of the vehicle, such as the yaw for a quadrotor. It is shown that for a quadrotor doing a 50 degree yaw change, the stability is greatly improved when the yaw axis is given very low priority.

Third, this thesis introduces the control of linear accelerations in all three axes with INDI. The controller does not need a complex model, but instead relies on a measurement of the acceleration. It is shown through a wind tunnel experiment, that the disturbance rejection properties, that were shown for the inner loop, carry over to the control of linear accelerations. It is also shown that the method can be applied outdoors with an off-the-shelve GPS receiver. Finally, a nonlinear method of calculating the input increment is derived, which provides only a slight improvement in the tracking of aggressive acceleration commands.

These three things are combined for the INDI control of hybrid MAVs. The result is a single, continuous INDI controller for the attitude, and a single, continuous INDI controller for the velocity of the vehicle. This is achieved by incorporating partial derivatives of the lift vector in the control effectiveness of the pitch and roll angles. Though no transition maneuver is explicitly defined, the transition follows implicitly from the increments in attitude and thrust that are calculated from desired acceleration changes. Further, as the control effectiveness of a hybrid MAV changes dramatically over the flight envelope, the control effectiveness is scheduled as a function of airspeed. When the airspeed is too low to measure, the pitch angle is used for this purpose. To prevent sideslip, a sideslip controller is included, where an estimate of the sideslip angle is obtained from the accelerometer.

Test flights show that the INDI inner and outer loop controllers are indeed capa- ble of controlling the attitude and the linear acceleration of the vehicle throughout the flight envelope, within the physical limitations of the vehicle. It is shown that the tracking of accelerations can make the vehicle naturally transition to forward flight and back, and fly in the stall region as necessary. Because of the abstraction that the INDI acceleration control provides, it is straightforward to follow a velocity vector field, for example one that guides the aircraft along a line.

The developed controller can be applied to different tailsitter MAVs with relative
ease, as the model dependency is low. The algorithm may even be applied to
different types of hybrid vehicles, such as quadplanes or tilt-wing aircraft, with
minor adjustments.