Feedforward Path Research Articles

Robustness of Gauss—Newton recursive methods: A deterministic feedback analysis

This paper provides a time-domain feedback analysis of the robustness performance of Gauss-Newton recursive methods that are often used in identification and control. These are recursive estimators that also involve updates of sample covariance matrices. Several free parameters are included in the filter descriptions while combining the covariance updates with the weight-vector updates. One of the contributions of this work is to show that by properly selecting the free parameters, the resulting filters can be made to impose certain bounds on the error quantities, thus resulting in desirable robustness properties (along the lines of H ∞ filter designs). It is also shown that an intrinsic feedback structure, mapping the noise sequence and the initial weight error to the a priori estimation errors and the final weight error, can be associated with such schemes. The feedback configuration is motivated via energy arguments and is shown to consist of two major blocks: a time-variant lossless (i.e., energy preserving) feedforward path and a time-variant feedback path. Emphasis is further given to filtered-error variants that give rise to dynamic time-variant feedback loops rather than memoryless loops. Such variants arise in IIR modeling.

Decoupling control of a static var compensator (SVC) with DC side voltage regulation and fast dynamics

This paper proposes a new control method of a static var compensator (SVC) with a three-level inverter, to control reactive power with fast dynamics while maintaining the DC side capacitor voltage constant. First, using the circuit DQ-tranformation, a general and simple model of SVC with a three-level inverter is obtained, and DC and AC analyses are carried out to characterize the open-loop system. Based on the transfer function matrix of the system, modulation index control for reactive power compensation and phase angle control for DC side voltage regulation are designed. By inserting a feedforward path to the voltage regulation loop the modulation index is successfully decoupled from the capacitor voltage, whereby fast dynamics are achieved. The experimental results confirm the theoretical analyses and control method.

Single-loop delta-sigma modulatorwith swing suppression

In the Letter a single-loop higher-order delta-sigma modulator with a new loop topology is presented. The loop-filter is composed of both feedback and feedforward paths. No active summing element is required in the modulator implementation. The modulator has low integrator output swings and a high input dynamic range, resulting in relaxed op-amp requirements even for low-voltage operation. Simulation results of a fourth-order modulator having the proposed loop filter are also given.

Optical adaptive neural networks with a ground glass for global random interconnections and local gain controls

TAG (training by adaptive gain) is a new adaptive neural network model for large-scale optical implementations. For single-layer neural networks withN input andM output neurons, the TAG model contains two different types of interconnections, i.e.MN global fixed interconnections and (N+M) adaptive gain controls. The training algorithm is based on gradient descent and error back-propagation, and is easily extensible to multilayer and/or higher-order architectures. For large-scale electrooptic implementation, the fixed global interconnections may be implemented by multifacet hologram, volume hologram or ground glass, and the adaptive gains by spatial light modulators (SLMs). The ground glass is more advantageous for random interconnections, with much higher diffraction efficiency and interconnection density. Both feed-forward signal and error back-propagation paths are implemented by a single ground glass, and adaptive learning has been demonstrated for heteroassociative memory and classifier.

A 3 V 12-55 MHz BiCMOS pseudo-differential continuous-time filter

The reduction of the supply voltage forces one to develop system and circuit solutions able to achieve the same performance previously obtained with higher supply voltage. In this paper, a second-order low-pass continuous-time filter operating at a 3 V power supply is presented. The prototype filter is implemented using a highly linear pseudo-differential transconductor. The input common-mode signal is canceled at the transconductor level using a feed-forward path. The output common mode voltage is controlled at the filter level using lossy integrators. A prototype cell has been realized in 1.2 /spl mu/m BiCMOS technology. The pole frequency can be tuned in the range 12-55 MHz. A THD of -40 dB is achieved for signals up to 1 V/sub pp/ at 10 MHz. The dynamic range is approximately 60 dB.

Modeling and control of magnet power supply system with switch-mode ripple regulator

This paper presents the dynamic performance analysis and closed-loop design of a new power supply system for a high-current magnet load. This new system introduces a series switch-mode ripple regulator (SMRR) into a phase-controlled rectifier to improve system dynamic response and to reduce the harmonic content in the magnet current. In this paper, the new system control configuration is explained. The small signal dynamic model is established. The function of the SMRR as a speed-up feed-forward path during dynamic transients is assessed. The criteria of the closed-loop regulator design are identified. The theoretical analyses are accompanied by computer simulation and experimental results. The systematical design procedure is outlined by an example which offers better performance than ever reported on high-current magnet power supplies in terms of fast dynamic response (closed-loop bandwidth several kHz) and low harmonic content (less than 10 ppm) in the magnet current.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

The solution of diagonal decoupling problem by dynamic output feedback and constant precompensator: the general case

In this paper, the solution to diagonal decoupling problem via dynamic output feedback and constant precompensator is presented in the general case. The literature contains a number of particular solutions for this problem. These solutions are valid, for instance, for square transfer matrices or they employ restricted dynamic output feedback schemes. Here, we start with a nonsquare transfer matrix and consider a dynamic output feedback scheme where there is a proper compensator in the feedback loop accompanied by a constant gain (precompensator) in the feedforward path. In this respect, the problem being considered is the dynamic output feedback version of the well-known Morgan's problem. The approach taken involves the characterization of open-loop diagonalizers which admit a dynamic output feedback realization. The key concepts involved in the solution are the generalized version of diagonal causality degree dominance and the set of attainable infinite zero orders. An example is also included to illustrate the main result. >

Minimax LQ exact and inexact model matching problems

A mixed H 2 H ∞ optimal problem is presented for both disturbance attenuation and model mismatching minimization. Its solution is in the form of a two-degrees-of-freedom controller, whose feedback and feedforward paths can be independently designed. If the conditions for exact model matching are not met, an inexact model matching problem is addressed. This solution provides an effective model matching design methodology for non-minimum phase plants.

High sample rate array architectures for median filters

Presents high sample rate semi-systolic array architectures for computing 1D and 2D nonrecursive and recursive median filters. A high sample rate is obtained by pipelining the computations in each processor. Although the nonrecursive filters are pipelined by placing latches in the feedforward paths, the recursive filters are restructured to create additional delays in the feedback paths, and then pipelined using the delays as latches. >

Model reference adaptive inverse control of a single link flexible robot

A new model reference adaptive inverse control algorithm is proposed for the end point position control of flexible manipulators. This controller forms a series model reference adaptive system where an estimated system inverse is placed in the feed forward path to provide an inverse control through regulation of controller parameters. It differs from other model reference control strategies in that a signal synthesis loop is included to compensate for the estimation error of the inverse controller and the output tracking error, and to improve stability. The stability of the overall control system is shown to be guaranteed using hyperstability theory. The proposed control scheme is applied to control the tip position of a single link flexible robot. A non-dimensionalized model is developed. Its zero-order hold equivalence in discrete time domain is described by an auto-regressive and moving average form. A lattice filter is utilized as a deconvolution filter. The time-delay behavior from the driving torque to the robot tip position due to robot structural flexibility is investigated.

A marionette-based strategy for stable movement

A strategy is developed for stable movement of a marionette under a system of unidirectional muscle-like actuators. It is shown that the strategy provides positive forces and positive inputs to the actuators that can be made analogous to monotonic function of the firing rate of natural muscles. The strategy requires more pairs of actuators than the degrees of freedom of the system-hence a need for synergistic actuators. In contrast to string-pulled marionettes where the length of the string is controlled, the tension of the string is controlled. Consequently the actuators are supplied with spindle-like position and velocity sensors and with independent input signals from higher control centers analogous to gamma inputs in living systems. Furthermore, ideal transmission delays are included in the feedforward and feedback paths to imitate neural transmission delays in living systems. To test the control strategy a three-link ten-actuator system is modeled and its stability, point to point movement, and tremor-like oscillations are shown by simulation. >

Exact and inexact LQG model matching problems

The design of discrete-time LQG (linear-quadratic-Gaussian) controllers that yield closed-loop systems that match a provided model is addressed. The solution is in the form of a two-degree-of-freedom controller having feedback and feedforward paths that independently provide the optimal actions for disturbance rejection and model matching. Exact and approximate model matching cases are discussed in a unified fashion. >

Stability of a DPCM transmission system with an order t predictor

Bounded-input-bounded-output (BIBO) stability is investigated for a DPCM transmission system. For the input signal, the output is calculated by a nonlinear feedback loop. The quantizer characteristics can be taken as continuous in the feedforward path. The quantizer exhibits a saturation level and has a linear part with a variable slope greater than 1. In the feedback path, the linear filter R is recursive. The stability of R is sufficient to ensure the stability of the DPCM system. The case of an order-1 predictor has been investigated previously. In this work, the stability of an order-t predictor is considered. The main result is that the stability of R is not necessary. In the second-order case, the shape of the stability domain in terms of the prediction coefficients is investigated.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Relationship between Model-Following Servo Control System and Two-Degree-of-Freedom Control System

We can recognize that both the model-following servo control system and 2-degree-of-freedom control system comprise the stable closed loop for the controlled object, the model, and the feedforward path from the model to the closed loop. In spite of the structural similarities, they are considered to be different systems. Since the mathematical tools used for the formulations are the state equation and the transfer function matrix, respectively, the approaches to solving the problem are different. However, the basic structures in both control systems are the same and it may be inconvenient for us to consider the two system as entirely different ones. Hence, clarifying the relationship between the model-following servo control system and the 2-degree-of-freedom (d.o.f.) control system is a very significant issue in the case of application to an actual plant. In this paper, the design method of the model-following servo control system and 2-d.o.f. control system are examined, and their similarities and differences are clarified. Next, the convergence to the model given by the designer is compared using the frequency response. In the model-following servo control system, the dimension of the model was not previously taken into consideration. However adequate convergence to the model can be shown by using numerical examples, if the dimension of the model is restricted in them.

Reactive Operations Scheduling for Manufacturing Systems

In this paper we present a reactive operations scheduling scheme for manufacturing systems. The closed loop system consists of a simulated annealing based predictive scheduler and the manufacturing system model in the feed forward path, and a re-scheduler in the feedback path. The deviation of the sequenced operation timings from the scheduled timings, quantified by some cost, is used to trigger the re-scheduling loop.

Minimax Controllers for Exact and Inexact Mode Matching

Abstract The solution to a mixed H2/H∞ optimal control problem is presented which allows disturbance rejection and model mismatching error minimization. The optimal solution is in the form of a two-degrees-of-freedom controller, whose feedback and feedforward paths can be independently designed and provide the optimal actions for disturbance rejection and model matching. This solution provides an effective model matching design methodology for non-minimum phase plants.

Negative feedback control model of proximal convergence and accommodation

A comprehensive model has been developed to illustrate the interactions between the observer and the surrounding environment in the control of oculomotor responses to distance or 3-D space. Accommodation and vergence respond to both spatiotopic (body reference) proximal percepts and retinotopic (eye referenced) physical stimuli of blur and disparity. Both spatiotopic and retinotopic stimuli are derived respectively from perceptual and physical correlates of negative feedback for eye position. The spatiotopic and retinotopic stimulus errors are combined in the feedforward path and drive a common oculomotor controller which has a phasic-tonic organization. Spatiotopic and retinotopic stimuli are shown to be effective over complementary operating ranges. Perceptual spatiotopic errors of gaze provide optimal stimuli for near responses to large depth intervals whereas physical-retinotopic cues of blur and disparity provide quantitative information about small binocular fixation errors. Small dynamic variations of target distance are sensed both spatiotopically and retinotopically. Coarse and fine spatiotopic errors of gaze are processed differently. Large spatiotopic errors are sampled intermittently at the beginning of the near response, whereas small retinotopic position errors and spatiotopic velocity errors are sampled continuously throughout the near response. Former reports of empirically observed higher velocity of vergence responses to very large depth intervals is explained in terms of stimulus sampling modes rather than in terms of separate oculomotor control mechanisms. The model demonstrates a complementary function of top-down spatiotopic cues, which are used to initiate the near response, and bottom-up retinotopic cues, which are used to refine and complete the near response. Cross-couplings by vergence-accommodation and accommodative-vergence serve to coordinate the components of the near response when feedback from sensed response of one motor system (i.e. vergence) is more accurate than that of the other motor system (i.e. accommodation). The model presented here is concerned primarily with the near response mediated by accommodation and disjunctive eye movements and not by the independent vergence mediated by non-conjugate or yoked saccades of unequal amplitude.

Iterative inversion of neural networks and its application to adaptive control

An iterative constrained inversion technique is used to find the control inputs to the plant. That is, rather than training a controller network and placing this network directly in the feedback or feedforward paths, the forward model of the plant is learned, and iterative inversion is performed on line to generate control commands. The control approach allows the controllers to respond online to changes in the plant dynamics. This approach also attempts to avoid the difficulty of analysis introduced by most current neural network controllers, which place the highly nonlinear neural network directly in the feedback path. A neural network-based model reference adaptive controller is also proposed for systems having significant dynamics between the control inputs and the observed (or desired) outputs and is demonstrated on a simple linear control system. These results are interpreted in terms of the need for a dither signal for on-line identification of dynamic systems.

A 100-MHz 100-dB operational amplifier with multipath nested Miller compensation structure

A multipath nested structure which overcomes the bandwidth reduction in the conventional nested structure by appending an independent feedforward path for high frequencies is presented. At low frequencies, the opamp behaves as a three-stage nested Miller-compensated amplifier, while at high frequencies the opamp has the nature and bandwidth of a two-stage amplifier with simple pole splitting. In this fashion, the concept circumvents the classical DC-gain/high-frequency-performance dilemma. No pole-zero doubles occur, because the two regions of operation can be independently and closely matched, unaffected by variations in the parameters of the IC process in which the amplifier is fabricated. >

Stability of the DPCM transmission system

Bounded input-bounded output stability of the differential pulse code modulation (DPCM) transmission system is investigated. The output is calculated by a nonlinear feedback loop. In the feedforward path the quantizer characteristic can be taken as continuous. It exhibits a threshold and has a linear part with a variable slope p greater than 1. In the feedback path the linear filter R is recursive. Stability of R is sufficient but not necessary to ensure stability of the DPCM system. Conditions for stability of an order 1 predictor are determined. Comparison with the classical concept of stability for linear systems displays three specific features of the DPCM case: 1) the conditions on the prediction coefficient are less stringent: 2) the kind of stability is weaker; and 3) the negative case has a better stability than the positive case. As a function of the quantizer slope, the maximum prediction coefficient exhibits a maximum. The corresponding value p/sub opt/ delimits two ranges. For smaller slopes the quantizer threshold effect degrades the stability. At the optimum, the range of stability is the same with and without the quantizer threshold.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>