Abstract: The present invention is a classification method and apparatus for classifying an input into one of a plurality of possible outputs. The invention generates a feature vector representative of the input. The invention then calculates a distance measure from the feature vector to the center of each neuron of a plurality of neurons, where each neuron is associated with one of the possible outputs. The invention then selects each neuron that encompasses the feature vector in accordance with the distance measure. The invention then determines a vote for each possible output, where the vote is the number of selected neurons that are associated with each possible output. If the vote for one of the possible outputs is greater than all other votes for all other possible outputs, then the invention selects that possible output as corresponding to the input.

Abstract: A hybrid fuzzy logic/neural network prediction system and method is disclosed for predicting response times to service requests to a service provider. Data from a historical database of records including customer requests and weather information are input to the hybrid system. The data is filtered to reject faulty data entries and data not necessarily useful for predicting response times to service requests such as customer comments are eliminated. A backpropagation neural network operating in a supervised learning mode is employed to decrease the effects of the inherent system nonlinearities. The prediction error from the neural network is trained to make predictions within a predetermined error limit. The neural network generates a prediction configuration; i.e. a set of neural network characteristics, for every record per geographical area, time frame, and month. A fuzzy logic classifier is used for further data reliability.

Abstract: An apparatus and method for automatically generating and adjusting fuzzy reasoning, rules based on changes in the reasoning error. The apparatus includes a fuzzy reasoning section that performs fuzzy logic based on the fuzzy reasoning rules stored in the a rule memory. The parameters of the fuzzy reasoning rules are adjusted in a parameter tuning section based on the output of the fuzzy reasoning section and predetermined input and output data. A reasoning error calculation section calculates a reasoning error and a change in the reasoning error based on the results from the fuzzy reasoning section and the predetermined input and output data. The reasoning error calculation section also disables the parameter tuning section when the calculated reasoning error is less than a predetermined first threshold value.

Abstract: Methods are developed on a digital computer for performing work order scheduling activity in a dynamic factory floor environment, in a manner which enables scheduling heuristic knowledge from a scheduler to be encoded through an adaptive learning process, thus eliminating the need to define these rules explicitly. A sequential assignment paradigm incrementally builds up a final schedule from a partial schedule, assigning each work order to appropriate resources in turns, taking advantage of the parallel processing capability of neural networks by selecting the most appropriate resource combination (i.e. schedule generation) for each work order under simultaneous interaction of multiple scheduling constraints.

Abstract: A neural network system comprising input, intermediate and output layers interconnected through synapses, respectively is disclosed.Each layer is comprised of a plurality of spatial light modulator units each of which is comprised of a photoconductive layer sandwiched between electrodes and a light modulation layer electrically connected to the photoconductive layer of which the light transmittance varies according to a voltage applied thereto, wherein electric currents induced by light bundles incident to the photoconductive layer are summed to cause a change in the voltage to be applied to the light modulation layer according to which the light transmittance is varied dependently thereon.

Abstract: A method for determining an optimal trajectory and velocity for an open loop stepper motor driven robot (1) is disclosed, as is an open loop robotic system suitable for use in, by example, a rapid prototyping system (10). The method utilizes a deflection angle calculation of a maximal velocity for each vertex of a set of vertices that define a desired trajectory, a heap sort for globally ensuring that none of the vertices have an excessive velocity, and a vertex adding technique that ensures that the robot is performing straight line moves as rapidly as is possible.

Abstract: A fuzzy neuron device which applies the fuzzy theory and the neuron theory. The device comprises a plurality of synapses each having a membership function, a first minimum value computing circuit for extracting the minimum value signal in the device from the outputs of the plurality of synapses, an extension circuit which is a component of a second minimum value computing circuits, and an extension terminal for connecting a plurality of extension circuit to each other so as to produce the second computing circuit. A device assembly may be produced by connecting a plurality of the devices to each other. In the device assembly, the second computing circuit extracts the minimum value signal in the assembly from the plurality of minimum signals in the respective devices.

Abstract: An arm origin calibrating method for an articulated robot is capable of implementing highly accurate calibration without requiring special high-precision measuring devices. A round bar is mounted on a first arm and a round hole is provided in a second arm. A rotation angle of the second arm is detected when, with the first arm being fixed, the second arm is rotated in the first direction until the round bar comes in contact with the inner surface of the round hole. A rotation angle of the second arm is detected when, with the first arm being fixed, the second arm is rotated in the second direction reverse to the first direction until the round bar comes in contact with the inner surface of the round hole. An offset angle of the second arm is detected on the basis of the two detected rotation angles.

Abstract: Even if a stop position of a supplied workpiece is not accurate, in order which an online robot can accurately perform works such as welding and painting, through the use of an offline robot of the same type as the online robot, detecting three reference positions on a workpiece by a first optical three-dimensional position detection means and teaching can be performed. The first coordinate transform matrix A1 stores the coordinates of each of the positions for an operation on the workpiece as the taught contents, when it is assumed that the workpiece is at a first reference position.

Abstract: An on-line training neural network for process control system and method trains by retrieving training sets from the stream of process data. The neural network detects the availability of new training data, and constructs a training set by retrieving the corresponding input data. The neural network is trained using the training set. Over time, many training sets are presented to the neural network.When multiple presentations are needed to effectively train, a buffer of training sets is filled and updated as new training data becomes available. The size of the buffer is selected in accordance with the training needs of the neural network. Once the buffer is full, a new training set bumps the oldest training set off the top of the buffer stack. The training sets in the buffer stack can be presented one or more times each time a new training set is constructed.

Abstract: A membership function generator including a maximum value arithmetic circuit for selecting a maximum voltage of a plurality of first input voltages to output the selected maximum voltage and a minimum value arithmetic circuit for selecting a minimum voltage of a plurality of second input voltages to output the selected minimum voltage. The outputs of the maximum and minimum value arithmetic circuits are input to a differential amplifier circuit. The output of the differential amplifier circuit is used as a membership function. Either at least one of the first input voltage or one of the second input voltages forms the basis of an inference and the remaining input voltages are set to a predetermined reference voltage.

Abstract: Process for generating a trajectory for a robotized system comprising at least one member and making it possible to vary the speed of the movement in real time.The process consists of choosing a main movement, e.g. the translation movement of the member and all the other movements are dependent thereon. The main movement takes place on the trajectory broken down into a section having a speed ensuring the possibility of stopping the member at the end of the section following that which has been covered. The dependent movements are covered in synchronized manner with the main movement to the extent that this is possible.The invention can e.g. apply to the automatic trimming or deburring of parts having a complex shape. It is possible to envisage robotized systems formed from several independently mobile solids coordinated by the process, or several control dependency instructions.

Abstract: A neural network (100) has an input layer, a hidden layer, and an output layer. The neural network stores weight values which operate on data input at the input layer to generate output data at the output layer. An error computing unit (87) receives the output data and compares it with desired output data from a learning data storage unit (105) to calculate error values representing the difference. An error gradient computing unit (81) calculates an error gradient, i.e. rate and direction of error change. A ratio computing unit (82) computes a ratio or percentage of a prior conjugate vector and combines the ratio with the error gradient. A conjugate vector computing unit (83) generates a present line search conjugate vector from the error gradient value and a previously calculated line search gradient vector. A line search computing unit (95) includes a weight computing unit (88) which calculates a weight correction value.

Abstract: A Cardinal number extending circuit for varying a resolution of input signals within a fuzzy neuron includes a plurality of computing blocks, each of the computing blocks having a different one of a plurality of resolution levels with each of the resolution levels represented by a Cardinal number. The Cardinal number for an n-th one of said computing blocks is 2.sup.n-1 k, where k is a number corresponding to a base resolution level. A switch responsive to an external selection signal selects a computing block corresponding to one of the Cardinal numbers to generate a computed result having one of the resolution levels.

Abstract: Electronic circuitry in the form of input and output circuits simulate the functions of neurons of most conceivable types. The input circuit has a signal regulating circuit which automatically reduces or increases its output amplitude, based on prior experience, simulating reverberation and memory neurons, respectively. An output stage has an integrator and a threshold circuit with a possible plurality of different types of inputs applied thereto for generating an output which simulates the responsiveness of a neuron. Different combinations of input and output circuits are used to simulate different types of neurons which can then be assembled into neural networks.

Abstract: A digital 1-bit wide abstractor for generating an exemplary pattern from a digital input information E, the abstractor having an evaluation unit which performs a transfer function p and to whose input the digital input information E is fed. The abstractor further includes a first decision unit having a first input which is connected with an output of the evaluation unit, a second input to which the input information E is fed, and third and fourth inputs to which a first predeterminable set value and a second predeterminable set value, respectively, are fed, with the first decision unit implementing an activation state according to an activation function, the first decision unit providing at an output thereof an output information according to an output function.

Abstract: A data processing system includes a memory with instructions and data stored therein. A fuzzy spreadsheet is stored in memory and has a plurality of cells, preferably arranged in columns and rows. The processor, stored instructions and stored data comprise a controller that receives input data and stores representations in cells. These representations include representations of fuzzy values. Operations, such as arithmetic operations, are performed on the data stored in the cells, including the fuzzy values, to produce derived representations that are stored in the cells and may be fuzzy values or crisp numbers or text. To facilitate a user's appreciation of a particular fuzzy number, the spreadsheet selectively simultaneously displays two representations of a fuzzy value, namely, a centroid and a graph.

Abstract: A membership function observation device capable of reducing the number of observation terminals and reducing the time required for measurement and control so that the interrelationship among membership functions can be understood. Pairs of membership functions output from the adjacent twos of a plurality of membership function generator circuits are supplied to respective minimum value circuits to obtain respective minimum values. These minimum values are supplied to a first maximum value circuit to obtain a maximum value. In turn, all the membership functions are supplied to a second maximum value circuit to obtain the maximum value of the membership functions. The outputs of the first and second maximum value circuits are synthesized to observe all the original membership functions at the same time.

Abstract: The invention controls an under-actuated robot manipulator by first obtaining predetermined active joint accelerations of the active joints and the passive joint friction forces of the passive joints, then computing articulated body quantities for each of the joints from the current positions of the links, and finally computing from the articulated body quantities and from the active joint accelerations and the passive joint forces, active joint forces of the active joints. Ultimately, the invention transmits servo commands corresponding to the active joint forces thus computed to respective ones of the joint servos.The computation of the active joint forces is accomplished using a recursive dynamics algorithm.

Abstract: A neural network provides both linearly separable and non-linearly separable logic operations, including the exclusive-or operation, on input signals in a single layer of circuits. The circuit weights the input signals with complex weights by multiplication and addition, and provides weighted signals to a neuron circuit (a neuron body or soma) which provides an output corresponding to the desired logical operation.