Augmenting L1 Adaptive Control of Piecewise Constant Type to Aerial Vehicles
Sammanfattning: In aerial vehicle control design, the industrial baseline is to use robust control methods together with gain-scheduling to cover the full airspeed and altitude flight envelope. An adaptive controller could possibly add value by increasing performance while keeping robustness to deviation from nominal assumptions. In this thesis L1 adaptive control is studied and evaluated as it is applied to a pitch-unstable fighter aircraft. The recently developed L1 adaptive control method originates from aerospace adaptive control problems and achieves fast adaptation while robust stability to bounded plant parameter changes is claimed. Even though large adaptation gains create large and rapidly varying internal signals, the L1 adaptive controller output is limited in amplitude and frequency, since a low-pass filter directly at the output, is used to make the controller act within the control channel bandwidth. An L1 adaptive controller of piecewise constant type has been applied to a fighter aircraft by augmenting a baseline linear state feedback controller. Once some experience is gained, it is relatively straightforward to apply this design procedure because only a few controller parameters need tuning. To design an L1-controller for roll-pitch-yaw-motion of an aerial vehicle, a five-state reference system with desired dynamics was created and five bandwidths of low-pass filters were tuned. The L1-controller activates when the vehicle aided by the state feedback controller deviates from the reference dynamics resulting in better reference following. Load disturbance rejection was improved by the L1-controller augmentation. This comes at the cost of having high frequency control signals fed into the plant. The L1 adaptive controller is in its original design sensitive to actuator limitations and to time delays when compared to the baseline controller. Introducing nonlinear design elements corresponding to actuator dynamics (e.g. rate limits) makes tuning easier if such dynamics interfere with the reference system dynamics. Sensitivity to known time delays can be reduced using prediction in a state observer. With these additions to the design, the L1-controller augmentation can be tuned to achieve improved nominal performance and robust performance when compared to a typical aeronautical linear state feedback controller. This was verified by simulations using a high fidelity model of the aircraft. Use of feedforward can alleviate feedback and adaptive actions. Feedforward signals can be generated from reference models and corresponding models can also be used as reference models in adaptive control. A method for aerial vehicle reference model design was developed, that makes it possible to find reference models that scale to the present flight condition and vehicle configuration. In some situations the closed-loop system obtained by L1 adaptive control is equivalent to linear systems. The architectures of these systems were investigated. An effort was made to understand and describe what fundamental characteristic of L1 adaptive controllers make them suitable for aeronautical applications. With the L1-controller, performance and robustness was increased when compared to the baseline controller. It is possible to add L1-controller characteristics gradually to a linear state feedback design, which is something that this thesis recommends to aerospace industry.
KLICKA HÄR FÖR ATT SE AVHANDLINGEN I FULLTEXT. (PDF-format)