Robust auto-landing of fixed-wing UAVs using neuro-adaptive design

An innovative neuro-adaptive design philosophy is presented in this paper embedding a Sobolev norm based Lyapunov function for directional learning of the unknown function, which is capable of learning both the unknown function in the system model and its Jacobian. This facilitates fast learning (model adaptation) without much of transient effects. The updated model is then used in the framework of dynamic inversion to design the guidance (outer) loop as well as the control (inner) loop. Using this philosophy a robust adaptive nonlinear guidance and control design is presented for robust automatic landing. The performance of the proposed approach is successfully verified through numerous simulation studies using the six degrees-of-freedom (six-DOF) nonlinear model of a prototype UAV. All possible disturbance effects that arise in practice, namely modeling inaccuracies, wind disturbances and ground effect, have been considered in the simulation studies.