Data processing: artificial intelligence – Neural network – Learning task
Reexamination Certificate
2000-11-21
2004-07-20
Picard, Leo (Department: 2121)
Data processing: artificial intelligence
Neural network
Learning task
C711S149000
Reexamination Certificate
active
06766310
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to electronic circuits, and, more particularly, to a neuro-fuzzy network architecture having self-training or on-line learning capabilities.
BACKGROUND OF THE INVENTION
The invention particularly relates to a self-training neuro-fuzzy network architecture including at least one fuzzy microcontroller (fuzzyfier/defuzzyfier) dedicated for calculating fuzzy rules. The fuzzy microcontroller may be integrated monolithically in a semiconductor along with a non-volatile memory, for example. The invention also relates to a method of electronically controlling semiconductor-integrated electronic devices using self-training.
Of the currently available electronic devices integrated monolithically in a semiconductor, the fuzzy logic products typically fall in the category of processors, microcontrollers, or general purpose fuzzy co-processors. However, these products are unsuitable for operation in an on-line learning and self-training mode. For example, there exist no commercially available devices which include circuit portions adapted to process dedicated self-training instructions.
The unavailability of integrated devices with self-training processing features makes prior art approaches ill-suited to address problems requiring a capability to accommodate process variations and changing environmental conditions. Such is the case, for example, with the control of fuel injection systems in internal combustion engines and in other automotive applications.
The present invention is directed to a semiconductor integrated electronic device for enabling processing of predetermined instructions or data processing procedures by self-training. The technical problem addressed by the present invention is to provide a programmable device as described above which is of the neuro-fuzzy type.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a programmable integrated device of the neuro-fuzzy type with dedicated hardware appropriate for on-line learning capabilities. In essence, the invention provides a fuzzy neural system (hereinafter Fuzzy Neural Network or FNN) which uses the properties of neural networks to automatically learn the variations of a process to be controlled, as well as to adjust its behavior accordingly by adaptation of the fuzzy rule system parameters for increased efficiency throughout.
Based on this concept, the objects are achieved by a self-training neuro-fuzzy network including at least one fuzzy microcontroller dedicated to fuzzy rules computing, a non-volatile memory integrated monolithically with the at least one fuzzy microcontroller on a semiconductor and connected thereto, a microprocessor, a volatile memory, and a bus interconnecting the fuzzy microcontroller. The microprocessor, the volatile memory, and an arbiter circuit are connected to the bus. The arbiter circuit controls access to the volatile memory by the microprocessor and the fuzzy microcontroller.
More specifically, the self-training neuro-fuzzy network may also include a fuzzy co-processor connected between the fuzzy microcontroller and the microprocessor for performing fuzzy logic operations. The fuzzy co-processor may be a slave to the microprocessor. Furthermore, an arithmetic logic unit may be connected to the microprocessor, and the arithmetic logic unit may include a sequential machine including a plurality of internal registers.
Additionally, the volatile memory may be a dual port random access memory. The self-training neuro-fuzzy network may further include an interface connected between the volatile memory and the non-volatile memory for exchanging data therebetween. Also, an input/output (I/O) module may be connected to the microprocessor for interfacing with external peripherals.
The invention also relates to a method of electronically controlling an electronic device by self-training where the electronic device is monolithically integrated on a semiconductor and includes at least one fuzzy microcontroller for fuzzy rule computing and a non-volatile memory connected thereto. The method includes writing data from the non-volatile memory to a volatile memory, reading data from the volatile memory and executing predetermined sequences of instructions on the data, connecting an arbiter circuit to a bus interconnecting the fuzzy microcontroller, the microprocessor, and the memory unit, and activating a fuzzy co-processor with the arbiter circuit. The fuzzy co-processor is activated upon receipt and recognition of a fuzzy logic instruction from the volatile memory. Additionally, the fuzzy co-processor may be operated as a slave to the microprocessor.
REFERENCES:
patent: 5245695 (1993-09-01), Basehore
patent: 5425131 (1995-06-01), Basehore
patent: 5459816 (1995-10-01), Basehore et al.
patent: 4225758 (1994-02-01), None
patent: 0742516 (1996-11-01), None
DANIELA: (Digital Analog Neuro-fuzzy Implementation of Enhanced Learning Algorithms) AT '96—1st International Symposium Workshop on Neuro-Fuzzy Systems, Aug. 1996, Marcello Chiaberge.*
Software-Hardware System for the Design of Adaptive Fuzzy Digital Signal Processing Applications, Labib Sultan, IEEE Member; (1998) IEEE, pps. 461-464.*
Processor Core for Adaptive Fuzzy and DSP Applications; Labib Sultan, IEEE Member (1998) IEEE, pps. 1072-1075.*
A High Performance Fuzzy Inference Processor and Its Evaluation System, Yasuhiko Nitta, Ken-ichi Shimomura, Narumi Sakashita, Shinji Komori, and Kazuo Kyuma; (1995) IEEE, pps. 1613-1620.*
Colodro et al., A circuit for learning in fuzzy logic-based controllers, 1995 IEEE Symposium on circuits and systems (Cat. No. 95CH35771), Proceedings on ISCAS' 95-International Symposium on Circuits and Systems, Seattle, WA USA, Apr. 28-May 3, 1995, pp. 279-282, vol. 1, XP0021555745, 1995, New York, NY, USA, IEEE, USA ISBN: 0-7803-2570-2.
Sultan,Processor core for adaptive fuzzy and DSP applications, 1998 IEEE International Conference on Fuzzy Systems Proceedings. IEEE World Congress on Computational Intelligence, (Cat. No. 98CH36228), Anchorage, AK, USA, May 4-9, 1998, pp. 1072-1075, vol. 2, XP002155743, 1998, New York, NY, USA, IEEE, USA, ISBN: 0-7803-4863-X.
Biz et al.,Furnace temperature control with fuzzy microcontroller, SMCIA/99 Proceedings of the 1999 IEEE Midnight-Sun Workshop on Soft Computing Methods in Industrial Applications, Kuusamo, Finland, Jun. 16-18, 1999, pp. 75-79, XP002155744, 1999, Piscataway, NJ, USA, IEEE, USA, ISBN 0-7803-5280-7.
Cola Eusebio Di
Lavorgna Mario
Occhipinti Luigi
Holmes Michael B.
Picard Leo
STMicroelectronics S.r.l.
LandOfFree
Neuro-fuzzy network architecture with on-line learning... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Neuro-fuzzy network architecture with on-line learning..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Neuro-fuzzy network architecture with on-line learning... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3255007