In this paper we introduce a novel mode-sensing system for aircraft, called a mission segment identifier (MSI). Using a blend of soft-computing approaches, the MSI infers the current portion of a commanded mission from sensed aircraft state variables. This inference contributes situation awareness to a pilot-like intelligent control system for autonomous aircraft. A two-level architecture for the MSI has been developed using neural networks for the lower level and a fuzzy inference system for the higher level. The resulting MSI has been validated using data from simulated flights of a conceptual autonomous aircraft. The performance of the MSI in the presence of sensor noise and ambiguities has been studied. A new method to synthesize training data for the neural network used in the lower level of the MSI evolved as part of this exercise. A simple, yet effective, scheme to reduce uncertainty in the identification has also been developed and validated.
[1]
Vance Quinton Harral.
Genetic algorithms for tuning fuzzy membership functions in flight control software
,
1995
.
[2]
D. Smith,et al.
An improved artificial neural network flight mode interpreter
,
1998
.
[3]
T. H. Nguyen,et al.
A neural-network based inference engine for a general aviation pilot advisor
,
1997
.
[4]
J. H. Painter,et al.
Decision support for the general aviation pilot
,
1997,
1997 IEEE International Conference on Systems, Man, and Cybernetics. Computational Cybernetics and Simulation.
[5]
Donald Ward,et al.
An intelligent flight director for autonomous aircraft
,
2000
.
[6]
J. H. Painter,et al.
Hypertrapezoidal Fuzzy Membership Functions for Decision Aiding
,
1998
.