摘要:
A fuzzy system (10) evaluates sensor signals from the motor vehicle (2) and generates regulating signals for controlled systems (5), such as automatic gear box, active suspension, speed stabilisation, assisted steering, drive slip control. The fuzzy system (10) is connected to a neuronal network (12) that evaluates the sensor signals from the motor vehicle, as well as reference data from a list of running data of the motor vehicle, in order to optimise the rule base of the fuzzy system (10). The fuzzy system (10) generates signals on-line during driving that categorise the present driving situation, thus achieving a control that is intelligent, adaptive in time and dependent on the driving situation. The fuzzy system (10) and the neuronal network (12) contain each a classification system. Both classification systems may be reciprocally converted by a correspondence-generating bi-directional transformation.
摘要:
The invention relates to a continuously variable transmission. The continuously variable transmission according to the invention is designed in such a way that in operating conditions to be defined in more detail the transmission ratio is kept essentially constant, in particular for travelling through bends, on inclines and in case of an emergency stop. The constant transmission ratio is obtained by regulating the input r.p.m. of the transmission.
摘要:
A control system for an automatic transmission, which can easily achieve a variety of shift patterns according to the taste of a driver and the situations of a running road. The control system comprises: a neural network; and a mode switch means for switching a teaching mode and an automatic mode. The neural network learns in the teaching mode and neuro-computes to output a gear ratio in the automatic mode. Further comprised are: tolerance error compute means for computing a tolerance error on the basis of at least one of th propriety of a teaching gear ratio, the ambiguity of an input value and the ambiguity of the teaching gear ratio; pattern produce means for producing a learning pattern; and combined load correct means for correcting the coupled loads of the neural network on the basis of the pattern produce means.
摘要:
A control device for an automobile has at least one sensor (8) for detecting an operational state of the automobile. An output from the sensor is applied to a sample and hold circuit (11) which is adapted to provide a time displaced output to a hierarchical group of neural elements (13). The neural elements operate upon said time displaced signals by varying the weighting of the signals and/or threshold values of the neural elements in accordance with a desired output parameter. The output of the neural computer is applied to a control circuit (2) which is adapted to control an automobile actuator. The hierarchical group of neural elements are provided with numerous forms of different learning functions so that the desired output parameter is achieved. In one such embodiment (Figure 32) desired parameters are applied to said neural computer and to a comparator, signals derived from sensors associated with automobile actuators are also applied to said comparator whereby difference signals are outputted from said comparator to control the weighting and/or threshold levels of the neural elements. The invention may be applied to different functions such as controlling air/fuel ratio, ignition timing, knock control, O₂ sensing, suspension adjustment, and road surface detection.
摘要:
A control device for an automobile has at least one sensor (8) for detecting an operational state of the automobile. An output from the sensor is applied to a sample and hold circuit (11) which is adapted to provide a time displaced output to a hierarchical group of neural elements (13). The neural elements operate upon said time displaced signals by varying the weighting of the signals and/or threshold values of the neural elements in accordance with a desired output parameter. The output of the neural computer is applied to a control circuit (2) which is adapted to control an automobile actuator. The hierarchical group of neural elements are provided with numerous forms of different learning functions so that the desired output parameter is achieved. In one such embodiment (Figure 32) desired parameters are applied to said neural computer and to a comparator, signals derived from sensors associated with automobile actuators are also applied to said comparator whereby difference signals are outputted from said comparator to control the weighting and/or threshold levels of the neural elements. The invention may be applied to different functions such as controlling air/fuel ratio, ignition timing, knock control, O₂ sensing, suspension adjustment, and road surface detection.
摘要:
Ein Verfahren und eine Vorrichtung zum Regeln eines Antriebsystems einer Maschine und/oder einer Anlage, bei dem Signale von Sensoren als Istwert-Meßdaten des Antriebsystems abgenommen, einer Datenerfassung zugeführt und in einem Berechnungssystem weiter verarbeitet werden, welches durch Algorithmen bereitgestellte Sollwerte der Meßdaten enthält und parameterbedingte korrigierte Stellsignale an ein Stellglied eines mit der Datenerfassung und dem Berechnungssystem ausgebildeten geschlossenen Regelkreises der Maschine und/oder der Anlage liefert, arbeiten mittels eines in das Berechnungssystem integrierten selbstlernenden Systems auf der Basis eines neuronalen Netzes, um eine adaptive Parametersteuerung des Regelkreises auf der Basis von abnormalen bzw. nicht eindeutig feststellbaren Meßdaten vorzunehmen.
摘要:
A control system for an automatic transmission, which can easily achieve a variety of shift patterns according to the taste of a driver and the situations of a running road. The control system comprises: a neural network; and a mode switch means for switching a teaching mode and an automatic mode. The neural network learns in the teaching mode and neuro-computes to output a gear ratio in the automatic mode. Further comprised are: tolerance error compute means for computing a tolerance error on the basis of at least one of th propriety of a teaching gear ratio, the ambiguity of an input value and the ambiguity of the teaching gear ratio; pattern produce means for producing a learning pattern; and combined load correct means for correcting the coupled loads of the neural network on the basis of the pattern produce means.
摘要:
A control section for controlling a gearshift position changeover mechanism of an automatic transmission includes a downshift determination section for determining whether or not a downshift on descending roads is required, and a downshift command section for outputting a downshift command signal when it is determined that downshift is necessary. The downshift determination section is provided with a learning correction section for effecting a learning correction of a determination reference based on a manipulation characteristic of a driver immediately after completion of downshift each time a downshift takes place.