Sharma-Flight-Path Angle Control via Neuro-Adaptive Backstepping.pdf

Flight-Path Angle Control via Neuro-Adaptive Backstepping Manu Sharma? and David G. Ward? Barron Associates, Inc. 1160 Pepsi Place Suite 300 Charlottesville, VA 22902 Abstract A method of controlling flight-path angle via neuro-adaptive backstepping is presented. Back- stepping is used to circumvent the matching con- dition, which can restrict feedback linearizing meth- ods. Additionally, an on-line multilayer neural net- work is used to provide robustness to aerodynamic uncertainties in the plant model. The network also simplifies the backstepping design because it obvi- ates the need to construct a regressor, and does not require an a-priori basis. Numerical simulation re- sults show the viability of this approach. Introduction Various forms of nonlinear control have emerged as enabling technologies for control of advanced flight vehicles.1—3 Offering both increases in per- formance as well as reduced development times by dealing with the complete dynamics of the vehicle rather than point designs, nonlinear control tools provide a great deal of design flexibility. Feedback linearization, in its various forms, is perhaps the most commonly employed nonlinear control method in this arena. However, the majority of the litera- ture on feedback linearizing controllers treats some combination of the aircraft’s angular rates (p, q, r), aerodynamic angles (α,β), or Euler angles (φ, θ,ψ) as the command variables. Using time-scale separa- tion arguments, the plant dynamics are partitioned into slow states and fast states, with the fast states used as virtual controls for the slow.4 Another re- striction on feedback linearization is the matching condition, which requires that all parametric plant uncertainties appear in the same equation of a state- space representation as the control.5,6 This is of significance in flight control where the plant aerody- namics always contain some degree of uncertainty. ?Research Scientist, Member AIAA. sharma@ ?Senior Research Scientist, Memb