Design and implementation of an adaptive controller for a quadcopter.

[摘要] ENGLISH ABSTRACT: The Solar Thermal Energy Research Group (STERG) from Stellenbosch Universityis attempting to reduce the cost of Concentrated Solar Power (CSP) plants. Introducing robotics into such CSP plants can assist in reducing thecost. An optimized calibration method currently being investigated by STERG is using a pair of quadcopters to calibrate the heliostats. The system requiresthat the quadcopter has a stable hover, despite the presence of disturbances.The aim of this project was to design and implement an advanced controller for a quadcopter. The advanced controller should improve on the existing controllerand ultimately allow for a more stable hover.Standard control laws have unique parameters that yield a certain response based on the model on which they are implemented. Changes in the model will result in changes in the response, yielding the need for new ideal parameters. Adaptive controllers have the advantage of reducing the number of control parameters to be tuned. Reducing the number of parameters can be beneficial,as obtaining the ideal parameters can become a time-consuming process.Model reference adaptive control (MRAC) is the control approach that is considered in this project. This approach has previously been implemented on a quadcopter by Achtelik [Adaptive Control of a Quadcopter in the Presence of large/complete Parameter Uncertainties, (2011)]. It was desired to implement the adaptive controller on the Pixhawk flight controller.The Pixhawk flight controller was chosen due to its capabilities when considering research projects. It also runs PX4 firmware which is part of an open source project. The designed controller should integrate well with the existing PX4 firmware to allow users still to be able to use the flight controller as before. In order to integrate the adaptive controller with the PX4firmware some modifications to the approach followed by Achtelik et al. (2011)was required. This report focuses on the implementation of MRAC in PX4 firmware. This required the use of quaternions in the control loop as opposedto the common Euler angles. The mixer was also extracted from the adaptive law. The mixer refers to the part of the controller which translates momentcommands to motor commands, according to the airframe being used.From simulations it could be seen that quaternions showed a significant improvement in reference tracking when it came to simultaneous pitch, roll andyaw inputs. The adaptive controller was first evaluated against other controllers in simulation before testing it in practice. In practical flight, it was again evaluated against other controllers. Specifically devised tests were evaluated to test the reference tracking and disturbance rejection of the different controllers. The adaptive controller showed the largest improvement, whencompared to the other controllers, in the disturbance rejection tests. Finally, an autonomous mission was flown with the newly designed adaptive controllerand also with the original PX4 controller. This showed successful integration of the adaptive controller with the existing firmware. An improvement in referencetracking for the adaptive controller, as opposed to the PX4 controller, was also found.

[发布机构] Stellenbosch University