Data processing: vehicles – navigation – and relative location – Vehicle control – guidance – operation – or indication
Reexamination Certificate
1999-03-11
2001-11-13
Cuchlinski, Jr., William A. (Department: 3661)
Data processing: vehicles, navigation, and relative location
Vehicle control, guidance, operation, or indication
C701S059000, C706S015000, C706S041000
Reexamination Certificate
active
06317658
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to a method, system and computer-readable medium for performing control in a vehicle and, more particularly, to a method, system and computer-readable medium for accurately determining control subsystem commands in real-time.
BACKGROUND OF THE INVENTION
Propulsion, aerodynamic, and other control subsystems in vehicles, such as vertical takeoff and landing aircraft, satellites, watercraft, and ejection seats, provide forces for controlling the operation of the vehicle. These control subsystems are controlled by controllers that distribute individual control effector commands to the control subsystems. These control distribution commands must provide timely and accurate commands in order for the control subsystems to keep the vehicle operating within predefined limits. For example, the controller in a vertical takeoff and landing aircraft must accurately supply control distribution commands to the propulsion system according to analysis of pilot entered commands, aircraft position and motion, and predetermined limits of the propulsion system, aircraft, and pilot. If the control distribution commands are untimely, jet engine response can lag or misinterpret pilot inputs to the point of causing dangerous vehicle oscillations. If the control distribution commands are inaccurate, jet engine total flow requirements will not be satisfied leading to engine flameout, degraded performance, or catastrophic failure.
Currently, controllers provide control distribution commands using iterative or multiple step algorithms implemented on digital microprocessors. This approach includes linear optimization or pseudo-inverse procedures that are inherently iterative in nature and have varying times of convergence. Iterative algorithms, when used for complex vehicle control, require extensive calculations and thus, have difficulty performing the required computations in real-time. When the iterative and multiple step algorithms cannot be executed in real-time, non-optimal approximate solutions are generated, thereby reducing the accuracy of the control distribution commands.
Another method of generating control distribution commands is a non-iterative linear technique, which uses mode switching when control limits are reached. This results in complex mode switching logic that is impractical to implement and verify. Another method implements multi-dimensional linearly interpolated data tables. This method requires excessive memory to store the data and is unable to execute the linear interpolation algorithms in real-time.
Current methods to design control systems that continue to provide controllability subsequent to control effector failures involve either adding components (redundancy) or reconfiguring the control laws (adaptive control) to rely on the unfailed components. Adding components increases system weight and cost. Adaptive control methods are slow to react to failures.
In summary, the present methods for providing control distribution commands require excessive microprocessor memory and complex mode switching, and are generally insufficient for providing real-time implementation.
The present invention is directed to overcoming the foregoing and other disadvantages. More specifically, the present invention is directed to providing a method, system and computer-readable medium for generating accurate control distribution commands in real-time.
SUMMARY OF THE INVENTION
In accordance with this invention, a neural network controller is trained off-line. The neural network controller is trained for all possible contingencies for the target, or on-line system. The data generated by the off-line, neural network controller is then used in the on-line system, thereby allowing the on-line system to operate in real-time. More specifically, in accordance with this invention, a method, system and computer-readable medium for generating control distribution commands for controlling a control subsystem of a vehicle is provided. The control subsystem includes at least one control effector. The system includes at least one sensor for sensing vehicle position and motion, an optional device for generating operator control signals, and control laws for generating desired vehicle forces/moments. Also included is a neural network controller for generating control effector commands for the at least one control effector based on the generated desired forces/moments, wherein said neural network controller was trained based on pregenerated vehicle control distribution data.
In accordance with other aspects of this invention, the vehicle control distribution data is generated off-line using an algorithm that determines control effector commands based on the desired vehicle forces/moments. The algorithm takes into account control subsystem nonlinearities and constraints, and calculates control effector commands that yield achievable vehicle forces/moments that deviate from the desired vehicle forces/moments in a proportional, prioritized, optimal, or other manner. Actuator limitations, flow balance, aerodynamic nonlinearities, rocket burn time, atmospheric effects, and nozzle area conversions are examples of control subsystem constraints and nonlinearities.
In accordance with still other aspects of this invention, the neural network controller is a unified or a decoupled structure with activation functions. The activation functions are sigmoidal, radial basis or linear functions.
In accordance with further aspects of this invention, the neural network controller is trained off-line using dynamic derivatives and a node decoupled extended Kalman filter technique or a gradient descent technique.
In accordance with yet other aspects of this invention, the neural network controller is implemented as an analog electronic circuit, a computer program, or an application specific integrated circuit.
In accordance with still further aspects of the invention, control effector commands are re-calculated to accommodate a loss of one or more effectors.
As will be readily appreciated from the foregoing summary, the invention provides a new and improved method, system and computer-readable medium for providing control distribution commands to a vehicle's control subsystems. Because command processing time is greatly reduced, excessive time delays in control distribution commands, and dangerous control subsystem and vehicle responses are avoided.
REFERENCES:
patent: 4721273 (1988-01-01), Trikha
patent: 4846421 (1989-07-01), Trikha
patent: 5469530 (1995-11-01), Makram-Ebeid
patent: 5493631 (1996-02-01), Huang et al.
patent: 5583771 (1996-12-01), Lynch et al.
patent: 5590218 (1996-12-01), Ornstein
patent: 5745653 (1998-04-01), Jesion et al.
Vian, J.L. and Visnevski, N.A., “Neural Networks for Nonlinear Control Distribution,” Third International Federation of Automatic Control Symposium on Intelligent Components and Instruments for Control Applications, Jun. 9-11, 1997, Annecy, France.
Christensen O'Connor Johnson & Kindness PLLC
Cuchlinski Jr. William A.
Donnelly Arthur D.
The Boeing Company
LandOfFree
Neurocomputing control distribution system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Neurocomputing control distribution system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Neurocomputing control distribution system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2584315