FPGA Implementation of Neural Network-Based AGPC for Nonlinear F-16 Aircraft Auto-Pilot Control: Part 2 – Implementation of Embedded PowerPC™440 with AGPC
Vincent Andrew Akpan,
Dimitrios Chasapis,
George Dimitriou Hassapis
Issue:
Volume 9, Issue 2, December 2022
Pages:
37-65
Received:
10 May 2022
Accepted:
9 June 2022
Published:
8 September 2022
DOI:
10.11648/j.ajesa.20220902.11
Downloads:
Views:
Abstract: The computational overload involved in the implementation of nonlinear model predictive control (NMPC) cannot be over-emphasized due to the double optimizations involved for the nonlinear model identification phase as well as the NMPC controller design phase. The computational burden becomes more time-critical for constrained multivariable control systems with relatively short sampling time. This paper presents a novel and comprehensive model-based design (MBD) approach for real-time closed-loop implementation of a version of NMPC referred here as adaptive generalized predictive control algorithmic co-processor (AGPC algorithmic co-processor) integrated with a well-designed embedded PowerPC™440 processor core on Virtex-5 FX70T ML507 FPGA (field programmable gate array) for the auto-pilot control of a nonlinear F-16 aircraft with a sampling time of 0.5 second. The result shows that the real-time closed-loop implementation of the neural network identification and the AGPC algorithms on the FPGA with embedded PowerPC™440 processor combined with an AGPC algorithmic co-processor at each sampling time is accomplished within 0.16502 microseconds (μs) when compared to the 6.1048 seconds obtained using Intel® Core™2 CPU personal computer for the control of the auto-pilot unit of a nonlinear F-16 aircraft. The demonstrated and validated model-based FPGA implementation techniques can be adapted and deployed for the real-time control of multivariable control systems having relatively short sampling time. The computation time and FPGA device utilization at each stage of the MBD implementation are also presented.
Abstract: The computational overload involved in the implementation of nonlinear model predictive control (NMPC) cannot be over-emphasized due to the double optimizations involved for the nonlinear model identification phase as well as the NMPC controller design phase. The computational burden becomes more time-critical for constrained multivariable control ...
Show More
Sentiment Mining and Aspect Based Summarization of Opinionated Afaan Oromoo News Text
Wegderes Tariku,
Million Meshesha,
Ashebir Hunegnaw,
Kedir Lemma
Issue:
Volume 9, Issue 2, December 2022
Pages:
66-72
Received:
17 August 2022
Accepted:
7 September 2022
Published:
19 September 2022
DOI:
10.11648/j.ajesa.20220902.12
Downloads:
Views:
Abstract: Studying the specific subject of opinion mining has been a popular research area as a means of overcoming the challenge of user-generated content on the web, which can be challenging to manually collect, comprehend, summarize, and analyze for decision-making. Even though there are three various levels at which opinion mining can be done, the detail and complexity of feature level opinion mining outweighs its disadvantages. The goal of this research is to provide sentiment mining and aspect-based opinion summaries of service reviews in Afaan Oromo for Oromia Radio and Television Organization (ORTO). 400 reviews in all were gathered and used for news-related purposes from ORTO. The model has five elements, including document inspection, pre-processing, aspect extraction, polarity detection, and aspect-based sentiment summary, as well as a bar chart to show aspect-based sentiment summation. Five different processes make up the model: document review, pre-processing, aspect extraction, polarity detection, and aspect-based sentiment summarization. A bar chart is also utilized to visually depict aspect-based opinion polarity. For positive classes, 90% precision and 87% recall are accomplished, while for negative classes, 87% precision and 89.7% recall are attained. The main issue identified in this study is that users tend to express their opinions in a context-based or indirect manner. They could express their negative feelings with pleasant words or the opposite. Therefore, more research is required before the algorithm will take context-based or semantic opinion mining into account.
Abstract: Studying the specific subject of opinion mining has been a popular research area as a means of overcoming the challenge of user-generated content on the web, which can be challenging to manually collect, comprehend, summarize, and analyze for decision-making. Even though there are three various levels at which opinion mining can be done, the detail...
Show More
