In 1991, the Aeronautics Directorate of NASA Ames Research Center became involved in the management of this project and redirected it under the more traditional "dynamical modeling and identification" paradigm. Independently of this effort, the Robotics Institute of Carnegie Mellon University (CMU) initiated its own start-up project, with the objective of gathering flight data on the old Yamaha R-50 helicopter, retrofitted with such modern avionics as GPS and INS.
The flight data gathered by Carnegie Mellon University consisted in the experimental frequency responses from the lateral stick, the longitudinal stick, the directional pedals, and the collective pitch to the longitudinal velocity u, the lateral velocity v, the vertical velocity w, the roll rate p, the pitch rate q, and the yaw rate r of the helicopter. This frequency response data was fed to the CIFER identification software of NASA to produce a 8-D linearized model, which did not include the flapping mode. Parallel to this experimental approach, a generic nonlinear helicopter model, called ARMCOP (for ARMy heliCOPter), was developed from "first principles." This model has 10 degrees of freedom (6 rigid body, 3 rotor flapping, and 1 rotor rotation). This generic model can be "tuned" to a particular helicopter by "tweaking" some free parameters. Next, a linearized ARMCOP model, trimmed for hovering, was tuned using mass, inertia, and other mechanical parameters of the R-50 as measured at CMU to produce frequency responses very close to those of the CIFER of the R-50 . This successful frequency response plot matching for hovering was a clear indication that the dynamics of the Yamaha R-50 helicopter became fairly well understood and that the available models were reliable.
Once reliable models became available, the design of inner loops for hovering and forward flight was approached using H-infinity techniques.
For more information about this first part of the project, contact Michael Frye at frye_michael@hotmail.com.
More recently, this project has taken a more theoretical turn and has focused on the following conceptualized identification problem: Given a complex mechanical system running, in an imprecisely known fashion, over the cotangent bundle of the configuration space, given a series of "trim points" around which the linearized dynamics is identified (using frequency response techniques), the question is whether there exists a global, nonlinear system such that its linearized dynamics around the various "trim points" match the experimentally identified linear systems? The answer is NO, unless some integrability conditions are satisfied. The usefulness of the integrability conditions is that they provide some kind of a check on the reliability of the identified data.
From a broader persepective, probably the most interested customer in the autonomous helicopter technology is the Marine Corps. The missions that are envisioned at this juncture for autonomous helicopters are reconnaissance and cargo. In 1998, the Kosovo conflict acted as yet another driving force behind the ARMY and the NAVY efforts on autonomous helicopters.