2018-02-07

bad things done by smart people

1. Performirng Target Classification
Using a Fuzzy Morphology Neural Network, (meaning the ai has a choice of, if it want to shoot you, it can or not, or even apologise then shoot you, idiot
Jennifer L. Davidson   (moron of the highest calibre)
Assistant Professor
Department of Electical Engineenng
Iowa State University
Abstract
This report details a neural network application to target classification using a Jifferent type of
neural network called a fuzzy morphological neural network. Neural networks are used mostly as patem
recognizers, and the main goal of this research was to give proof of concept that a simple form of this
new neural network could perform object classification. The type of d".ta used was the range data irOm
ladar data on tanks. A two-tiered approach was used on the data. First. a linear convolution with a fixed
template was performed. Then this data was input to a neural network for classification. The image
was divided up into zones, with the expectation that a partially occluded object couid still carry forward
enough information from fewer zones to the neural network to identify the object. The simplified neural
network was essentially a multilaver percepiron with backpropagation. so this network was used to train
the data. The results showed that with proper training, the network can perform very good classificauon.
The numerical results are given in this report. and substantiate our initial ideas. With the proof of concept
in hand, the next step is to develop training algorithms and network architectures which take advantage
of the complex range of parameters available in the network. This is the main topic of research cescribed
in a follow-up minigrant proposal.

PRELIMINARY MISSILE AUTOPILOT DESIGN
USING REACTION JETS AND AERODYNAMIC CONTROL
Mario Innocenti
Associate Professor
Department of Aerospace Engineering
Auburn University, AL 36849-5338
Abstract
The feasibility of combining traditional aerodynamic control with reaction
jets is studied, in the framework of missile autopilot design. The purposes of
reaction jets are to increase the missile turning capabilities in critical phases of flight,
as well as to help the traditional aerodynamic control in configurations where the
fin size is reduced because of limited storage volume. Due to the nonlinear
characteristics of both controller and airframe dynamics during fast maneuvers, a
control strategy based on variable structure systems is used. A control law is
synthesized for the pitch loop and some initial results are presented and compared
with a traditional controller based on linear quadratic technique



DEVELOPMENT OF CONTROL DESIGN METHODOLOGIES
FOR FLEXIBLE (HIGH ORDER) MISSILE SYSTEMS
WITH MULTIPLE HARD NONLINEARITIES
Armando A. Rodriguez
Assistant Professor
Deptartment of Electrical Engineering
Arizona State University
1 Abstract
The primary goal of this research endeavour has been to initiate a program directed at the development
of practical methods for designing robust full envelop autopilots for flez:ble missile systems with multiple
hard nonlineant:es. The methods developed will permit control engineers to systematically generate full
envelop controllers on the basis of very complex (high order) missile models.
The research has been driven by the ultimate goal of designing a full envelop autopilot for an Extended
Medium-Range Air-to-Air Technology (EMRAAT) Bank-to-Turn (BTT) missile with significant structural
modes. Of particular interest is the terminal phase of an air-to-air intercept during which control surfaces
may saturate in both position and rate. This is critical because structural modes may be excited and, more
importantly, missile destabilization may be induced. Assuring robust performance during such a high control
scenario has guided the path of this research.
Contents
1 Abstract 2
2 Motivation 3
2.1 Controlling Flexible (High Order) Missile Systems . . . ... ............... 3
2.2 Controlling Systems with Multiple Hard Nonlinearities ........................... 3
3 A Methodology for Infinite-Dimensional Systems 3
3.1 Introduction ........... .............................................. 3
3.2 Statement of Three Fundamental Problems .................................. 4
3.3 N*0 Mod -i Matching Problems ........ .................................. 7
3.3.1 iifinite-Dimensional Nf' MMP ..................................... 7
3.3.2 Sequences of Finite-Dimensional 7WO MMP's ............................ 9
3.4 Solution to 7il Sensitivity Problems ....... ............................... 11
3.5 Solution te NIo Mixed-Sensitivity Problems ....... ........................... 13
4 A Methodology for Systems with Hard Nonlinearities 15
5 Significance of Work 16
6 Summary & Directions for Future Research 17
7 Bibliography 18
8 Appendix: Control of a BTT Missile with Saturating Actuators 20
H-145

No comments: