Method of training a neural network for the guidance of a...

Details Method of training a neural network for the guidance of a... Method of training a neural network for the guidance of a...

1999-07-14

2003-09-30

Follansbee, John (Department: 2121)

Data processing: artificial intelligence

Fuzzy logic hardware

Fuzzy neural network

Reexamination Certificate

active

06629085

ABSTRACT:

BACKGROUND OF THE INVENTION
The invention relates to a method of training a neural network such that it will be able to guide a missile to a target.
Target tracking missiles have a seeker head which is able to detect, in an object scene, a target to be tracked. The seeker head and a signal processing unit, to which signals from the seeker head are applied, provide guidance signals. The guidance signals cause deflection of steering surfaces of the missile and, thereby, an angle of attack and a transverse acceleration such that the missile is guided on a collision course to the target. In prior art missiles, guidance of the missile is effected in accordance with the guidance law of “proportional navigation”. With this mode of guidance, the angular rate of the line of sight to the missile in inertial space is determined. The transverse acceleration is made proportional to this angular rate.
If the target has high maneuvering capacity, is, for example, a highly maneuverable fighter aircraft, a “schematic” guidance in accordance with the guidance law of proportional navigation will not suffice. The required guidance laws become highly non-linear, when all occurring situations are to be taken into account.
SUMMARY OF THE INVENTION
It is an object of the invention to provide improved guidance of target tracking missiles.
To this end, a neural network is trained to guide a missile to a target. According to one aspect, the method of training the neural network comprises the steps of representing a scenario of missile and target, transforming this scenario into slow-motion, simulating the flight of the missile to the target, a human pilot guiding the missile to the target, storing the pilot's behavior and the reaction of the missile resulting therefrom for a number of such simulated flights, re-transforming the data thus stored into real time, and training a guidance unit provided with a neural or fuzzy-neural network with the behaviour of pilot and missile re-transformed into real time.
Thus the missile is equipped with a “virtual human pilot” in the form of a correspondingly trained neural network. The neural network reacts in the same way as a human pilot sitting in the missile would react. As the human pilot is not able to react as quickly as the events during tracking of the target by the missile take place, the target tracking is, at first, simulated in slow-motion. The behavior of pilot and missile is recorded, converted into digital data and stored. Then the data recorded in slow-motion are re-transformed into real time. Thereby, a set of simulation data and pilot's reactions for the various simulated situations is obtained. A neural or fuzzy-neural network is trained therewith. Thereby, the neural or fuzzy-neural network is “cloned” with the pilot's behavior and knowledges, and then behaves like a human pilot sitting in the missile.
According to another aspect of the invention, the solution of the non-linear guidance problem is computed in analytical form. Numerical solutions of a number of flights to a target are generated. The behavior of a pilot and of the missile is determined for a number of flights of the missile to a target by simulation. A neural or fuzzy-neural network is “cloned” with the knowledge about the guidance of the missile to the target, as obtained by the preceding steps.
Then, the neural or fuzzy-neural network, for the guidance of the missile, makes not only use of the human pilot's experience but, in addition, of analytically or numerically obtained knowledge about the behavior of the missile.


REFERENCES:
patent: 5259064 (1993-11-01), Furuta et al.
patent: 5751915 (1998-05-01), Werbos
patent: 5768122 (1998-06-01), Motoc
patent: 5924085 (1999-07-01), Werbos
patent: 6169981 (2001-01-01), Werbos
Design of Advanced Guidance Law against High Speed Attacking Target, Chun-Liang Lin and Yung-Yue Chen, Proceeding National Science Council ROC(A), vol. 23, No. 1, (1999) pps. 60-74.*
Bayesian Belief Update in Antiair Defense, Sanguk Noh and Piotr J. Gmytrasiewicz; (1997) pps. 1-5.*
Optimal Controller Approximation Using Neural and Fuzzy-Neural Networks, Michael Niestroy; IEEE, (1996) pps. 486-491.*
THESIS; Physical Based Toolkit For Real-Time Distributed Virtual World, Henry Tong Ong, (Sep. 1995), Naval Postgraduate School, Monterey, California, pps. 1-54.*
Optimistic Real-Time Simulation; Kaushik Ghosh, Richard M. Fujimoto, and Karsten Schwan; (Jun. 28, 1995), pps. 1-17.*
Adaptive Traget State Estimation Using Neural Networks, P. K. Menon and V. Sharma; Optimal Synthesis, (1999), pps. 1-6.*
Computing, Research, and War: If Knowledge is Power, Where is responsibilty? Jack Beusmans and Karen Wieckert, Communications of trhe ACM, (Aug. 1989), vol. 32, No. 8, pps. 939-947.

Affiliated with

Also associated with

Rating

Method of training a neural network for the guidance of a... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Method of training a neural network for the guidance of a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of training a neural network for the guidance of a... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3025491