Loading...
Adaptive Neural Control of Gimbaled Laser Targeting System
Giorgi, Salvatore
Giorgi, Salvatore
Citations
Altmetric:
Genre
Thesis/Dissertation
Date
2014
Advisor
Committee member
Group
Department
Electrical and Computer Engineering
Subject
Permanent link to this record
Collections
Research Projects
Organizational Units
Journal Issue
DOI
http://dx.doi.org/10.34944/dspace/2903
Abstract
Space-based solar power is a better alternative to the ground-based solar system, because of its round-the-clock availability. Accurately pointing the laser from space to a ground receiver can increase transmission efficiency and decrease the receiving area. A superior pointing performance requires an accurate system model and controller. In this thesis, we investigate Adaptive Neural Control for a laser targeting application. A two-axis Gimbaled Laser Target System (GLTS) is used as hardware test bench of the space-based solar power transmission system. The Adaptive Neural Control (ANC) system, first proposed by D.C. Hyland, is a neural control system within a Model Reference Control (MRAC) architecture. It is composed of five separate neural networks, two of which are used to replicate an unknown plant, while the remaining three are used to control the plant's output to match that of an ideal reference system. The system has been successfully used in hardware such as the NASA / LaRC Mini-MAST testbed and the ASTREX testbed at Air Force Philips Laboratory. It has been shown to be very effective in terms of robustness, fault tolerance, and optimality. The objective of this research is to apply the ANC system to the problem of pointing and tracking the line of sight of a GLTS. A software model of the ANC system is built using Matlab / Simulink. We then simulate control of a linear stochastic model of our GLTS test bed and compare the ANC system's performance to that of a Proportional Integral Derivative (PID) controller. Next, we consider a separate nonlinear stochastic model of a two-axis gimbal, and consider the problem of platform stabilization using the ANC system. Next, we examine the ANC system's resiliency, defined in terms of how the controller maintains operational normalcy in response to anomalies, both unexpected and malicious. These anomalies will be in the form of added latencies, plant parameter changes, false data injection, and sensor data alteration. We simulate the attacks on the GLTS model and determine the system's resiliency to each attack through four metrics. These metrics are recovery time, performance degradation, protection time, and degrading time. We then compare these results to that of a PID controller subjected to the same attacks. Finally, the ANC system software model is translated to a fixed point hardware model for implementation on a Field Programmable Gate Array (FPGA) using Xilinx / System Generator. This software to hardware translation considers special attention to floating point to fixed point conversion, division, representing nonlinear neural functions, and hardware resource allocation. System replication and control simulations are run for the linear stochastic GLTS model. These simulations include ``hardware in the loop" simulations, where the actual FPGA is used within the Matlab simulation. Simulations show that the software model of the ANC system is able to replicate and control the linear GLTS model as well as a separate nonlinear gimbal model in the presence of process and measurement noise, with no prior modeling information. Additionally, the simulations demonstrate the resiliency of the ANC system when exposed to attacks, in terms of recovery time, performance degradation, and degrading time. Lastly, hardware simulations confirm the ANC system's ability to control the GLTS model using the FPGA.
Description
Citation
Citation to related work
Has part
ADA compliance
For Americans with Disabilities Act (ADA) accommodation, including help with reading this content, please contact scholarshare@temple.edu