System Identification and Model-Based Control of Quadcopter UAVs

As control systems become more sophisticated, more accurate system models are needed for control law design and simulation. In this research, a nonlinear dynamic model of a quadcopter UAV is presented and model parameters are estimated off-line using in-flight experimental data. In addition, a model-based classical control law for the quadcopter UAV is designed, simulated, and then deployed in UAV flight tests. The intent of this research is to identify a model which may be simple enough to easily use for control law design, and accurate enough for simulation. In addition, a model-based classical control law is designed to for flight control. The parameters of the nonlinear dynamic model are estimated with the Linear Least Squares Error method. In-flight disturbances are introduced in flight tests to ensure frequency rich data. The performances of different models are compared using validation flight test data to select an accurate model. This model is used as the simulation model and the design model. Model-based control law design techniques are used to create a flight control law which provides good performance both in the simulator, as well as when deployed to the quadcopter. To perform these tests, the Real-Time - Marseille Grenoble Project software is used for the creation of ground station programs and flight control algorithms in Simulink. This test environment integrates a VICON camera systems, QuaRC Real Time system, a 3DR APM 2.6 micro-controller unit, and a Gumstix Overo AirSTORM micro-controller unit to create a low-cost quadcopter research platform