Intelligent Nonlinear Control of Remotely Piloted Model Scale Helicopters

An intelligent nonlinear control methodology, based on an innovative combination of feedback linearization and artificial neural networks, is to be adapted to control remotely piloted helicopters. A mapping is derived which transforms a nonlinear dynamic model of the plant into a linear, time invariant space. A single point linear controller design is then carried out in this linear space. The derived control is then mapped back into the nonlinear space using an inverse transformation. An artifical neural network can be trained prior to flight to realize the inverse mapping in real time. A second network can be trained in flight to learn, and correct for, the vehicle characteristics not captured in the dynaic model used to derive the control. The feasibility and value of this approach is to be demonstrated by evaluation of a controller design for the current US Army/NASA X-Cell model-scale helicopter using a high fidelity dynamic simulation. The design will include optional position feedback loops which can be used to track a prescribed trajectory. Once the computational burden of the control algorithm has been quantified, conceptual design of flight worthy hardware will begin. Detailed design, construction, test and evaluation will be carried out in Phase II.