51. Modelling, control and construction of tricopter unmanned aerial vehicles
- Author
-
Abara, Daniel and Lanzon, Alexander
- Subjects
tricopter ,cooperative control ,UAV ,consensus ,system identification ,trirotor ,quaternion ,sliding mode ,model predictive control ,control ,negative imaginary - Abstract
This thesis deals with the development and control of low-cost single-tilt and multi-tilt tricopter aerial vehicles. Tricopter UAVs have been shown to be more agile and manoeuvrable offering more advantages than other multicopters like the quadcopter for example. The dynamic models for both tricopters are derived from first principles and experimental data is used to obtain the actuator constants. In the case of the single-tilt tricopter, a control allocation algorithm is also proposed to solve the problem of the number of control inputs being more than the number of actuators since the single-tilt tricopter has only four actuators (3 rotors and 1 servo) and a greater number of control inputs (forces and torques). A cascaded-PID control scheme is then used to stabilize the single-tilt tricopter in hover mode. The simulation results yield realistic control inputs and the outputs have acceptable performance. The feasibility of the proposed scheme is then validated with some experiments on the developed tricopter platform in hover. For the multi-tilt tricopter, a Quaternion Feedback Control (QFB) scheme is proposed which uses unit quaternions to represent the attitude dynamics in order to avoid gimbal lock which occurs when the pitch angle approaches plus or minus 90 degrees if using Euler angles. Also, a linear Model Predictive Control (MPC) scheme which uses the Linear Parameter Varying model of the tricopter is proposed. These control techniques are tested and compared using simulations in Matlab/Simulink. The feasibility of achieving independent position and attitude control, with possibility of translating in the longitudinal and lateral directions without changing the multi-tilt tricopter's attitude, is shown in simulation and demonstrated in experiments on the in-house tricopter. The hardware implementation of this concept is achieved by developing an algorithm using the PX4 framework which allocates the lateral and longitudinal forces via mapped transmitter knobs. Finally, this thesis also proposes a robust leader-following formation control scheme for a class of multi-agent systems that can be modelled as a group of networked closed-loop linearized multi-tilt tricopter agents utilizing Negative Imaginary (NI) theory. A continuous time subspace identification method for NI systems based on the Laguerre filter is proposed, which guarantees that the resultant model is NI. Sliding Mode Control (SMC) is used to linearize the tricopter system in closed-loop having six inputs and outputs instead of the more common feedback or Jacobi linearization methods. The usefulness of the identification algorithm is shown via the identification of an NI model for the SMC-linearized tricopter (inner-loop). The leader-following formation problem is formulated as an asymptotic tracking problem of a distributed Strictly Negative Imaginary (SNI) plus Very Strictly Passive (VSP) system being cascaded with a network of closed-loop linearized multi-tilt tricopter agents. An in-depth simulation case study is performed on a formation tracking mission for a group of six SMC-linearized multi-tilt tricopters and the results show that the tricopter agents achieve consensus tracking and leader-following group formation tracking.
- Published
- 2022