A Nonlinear Output Regulation Perspective on

Autonomous Helicopter Landing

 

Professor Andrea Serrani

 

Department of Electrical Engineering

The Ohio State University

 

Autopilot design for helicopters is a challenging testbed in nonlinear feedback design,  due to the nonlinearity of the dynamics and the strong coupling between the forces and torques produced by the vehicle actuators. A helicopter is, in general, an underactuated mechanical system, that is, it possesses more degrees of freedom than independent control inputs. This accounts for the presence of a nontrivial internal dynamics when feedback linearization techniques are applied. In fact, the resulting zero-dynamics are critically stable, which means that the system exhibit a non-minimum phase behavior which complicates tremendously the synthesis of nonlinear control laws. Moreover, the model may be affected by large uncertainties and unmodelled dynamics. In this talk, we address the design of an internal-model based autopilot for a helicopter, capable to let its vertical position follow an exogenous reference signal, while stabilizing the horizontal and

lateral position and the vehicle attitude to a constant configuration. The exogenous reference is not supposed to be measured, rather only the error and its first derivative are accessible to the controller. The reference signal is given as a sum of a constant term and a fixed number of sinusoidal signals of unknown frequency, amplitude and phase.  This scenario arises when one considers the problem of letting the helicopter land autonomously on an oscillating platform, as, for instance, a ship subject to wave-induced oscillations. A similar problem has been previously considered and solved for a simplified model of a VTOL aircraft. With respect to the former,  however, the present case is more challenging, due to the higher complexity of the vehicle dynamics which renders the stabilization onto the desired trajectory a difficult task. We propose a solution which combines recent results on nonlinear adaptive regulation and robust stabilization of systems in feedforward form by means of saturated controls. The focus of the talk is on the design of an adaptive internal model and on the peculiar robust stabilization technique employed. Due to the intrinsic robustness of the method, we expect the controller to perform satisfactorily despite the effect of parametric uncertainties and unmodeled dynamics, as suggested by nonlinear simulation on a full-order dynamic model.

 
Friday, October 4, 2002

3:30 – 4:30 p.m.

1500 EECS