Showing papers in "IEEE Control Systems Magazine in 1997"

Modeling piezoelectric stack actuators for control of micromanipulation

Abstract: A nonlinear lumped-parameter model of a piezoelectric stack actuator has been developed to describe actuator behavior for purposes of control system analysis and design, and, in particular, for microrobotic applications requiring accurate position and/or force control. In formulating this model, the authors propose a generalized Maxwell resistive capacitor as a lumped-parameter causal representation of rate-independent hysteresis. Model formulation is validated by comparing results of numerical simulations to experimental data. Validation is followed by a discussion of model implications for purposes of actuator control.

Controlling buildings: a new frontier in feedback

Abstract: The protection of civil structures, including their material contents and human occupants, is without doubt a worldwide priority of the most serious importance. Such protection may range from reliable operation and comfort, on the one hand, to survivability on the other. Examples of such structures leap to one's mind, and include buildings, offshore rigs, towers, roads, bridges, and pipelines. In like manner, events that cause the need for such protective measures are earthquakes, winds, waves, traffic, lightning, and-today, regrettably-deliberate acts. Indications are that control methods will be able to make a genuine contribution to this problem area, which is of great economic and social importance. In this article, we review the rapid developments which have been occurring in the area of controlled civil structures, including full-scale implementations, actuator types and characteristics, and trends toward the incorporation of more modern algorithms and technologies.

Nonlinear design of active suspensions

Abstract: This article develops a new nonlinear backstepping design for the control of active suspension systems, which improves the inherent tradeoff between ride quality and suspension travel. The novelty is in the use of a nonlinear filter whose effective bandwidth depends on the magnitude of the suspension travel. This intentional introduction of nonlinearity, which is readily accommodated by backstepping, results in a design that is fundamentally different from previous ones: as the suspension travel changes, the controller smoothly shifts its focus between the conflicting objectives of ride comfort and rattlespace utilization, softening the suspension when suspension travel is small and stiffening it as it approaches the travel limits. Thus, our nonlinear design allows the closed-loop system to behave differently in different operating regions, thereby eliminating the dilemma of whether to use a soft or stiff suspension setting. The improvement achieved with our design is illustrated through comparative simulations.

300 years of optimal control: from the brachystochrone to the maximum principle

Abstract: An historical review of the development of optimal control from the publication of the brachystochrone problem by Johann Bernoulli in 1696. Ideas on curve minimization already known at the time are briefly outlined. The brachystochrone problem is stated and Bernoulli's solution is given. Bernoulli's personality and his family are discussed. The article then traces the development of the necessary conditions for a minimum, from the Euler-Lagrange equations to the work of Legendre and Weierstrass and, eventually, the maximum principle of optimal control theory.

Robust control prevents car skidding

Abstract: The author discusses a system of robust unilateral decoupling of car steering dynamics. Its effect is that the driver has to concern himself much less with disturbance attenuation. The important quick reaction to disturbance torques is done by the automatic feedback system. The yaw dynamics no longer interfere with the path-following task of the driver. The safety advantages have been demonstrated in experiments with a test vehicle. By empirical improvements, we have modified the controller such that it preserves the robust decoupling advantages for the first 0.5 seconds after a disturbance and then returns the steering authority gradually back to the driver.

Variable structure control of shape memory alloy actuators

Abstract: A shape memory alloy (SMA) actuator consisting of a number of thin NiTi fibers woven in a counter rotating helical pattern around supporting disks is first described. This structure accomplishes a highly efficient transformation between force and displacement overcoming the main mechanical drawback of shape memory alloys, that being limited strain. Time domain open loop experiments were then conducted to determine the intrinsic properties of the actuator. From these experiments and from the knowledge of the underlying physics of SMAs, a multiterm model, including linear and nonlinear elements, was proposed. After further investigation and simulation, it was found that most of these complexities did not need to be considered in order to explain the reported results, and that the model could be reduced to that of a single integrator. A variable structure controller was then applied to a pair of antagonist actuators. The feedback switches between the two actuators according to the sign of the displacement error. A further improvement was added to compensate for known gross nonlinearities by modulating the current magnitude in a discrete manner as a function of the state space position. It was therefore possible to realize smooth and robust control with very little cost in complexity.

Control architectures for autonomous underwater vehicles

Abstract: Autonomous underwater vehicles (AUVs) share common control problems with other air, land, and water unmanned vehicles. In addition to requiring high-dimensional and computationally intensive sensory data for real-time mission execution, power and communication limitations in an underwater environment make it more difficult to develop a control architecture for an AUV. In this article, the four types of control architectures being used for AUVs (hierarchical, heterarchical, subsumption, and hybrid architecture) are reviewed. A summary of 25 existing AUVs and a review of 11 AUV control architecture systems present a flavor of the state of the art in AUV technology. A new sensor-based embedded AUV control system architecture is also described and its implementation is discussed.

Control for slosh-free motion of an open container

Abstract: This article describes two methods for controlling the surface of a liquid in an open container as it is being carried by a robot arm. Both methods make use of the fundamental mode of oscillation and damping of the liquid in the container as predicted from a boundary element model of the fluid. The first method uses an infinite impulse response filter to alter the acceleration profile so that the liquid remains level except for a single wave at the beginning and end of the motion. The motion of the liquid is similar to that of a simple pendulum. The second method removes the remaining two surface oscillations by tilting the container parallel to the beginning and ending wave. A double pendulum model is used to determine the trajectory for this motion. Experimental results of a FANUC S-800 robot moving a 230 mm diameter hemispherical container of water are presented.

A systematic classification of neural-network-based control

Abstract: Successful industrial applications and favorable comparisons with conventional alternatives have motivated the development of a large number of schemes for neural-network-based control. Each scheme is usually composed of several independent functional features, which makes it difficult to identify precisely what is new in the scheme. Help from available overviews is therefore often inadequate, since they usually discuss only the most important overall schemes. This work breaks the available schemes down to their essential functional features and organizes the latter into a multi-level classification. The classification reveals that similar schemes often get placed in different categories, fundamentally different features often get lumped into a single category, and proposed new schemes are often merely permutations and combinations of the well-established fundamental features. The classification has two main sections: neural network only as an aid; and neural network as controller.

A new matrix model for discrete event systems: application to simulation

Abstract: Simulation schemes for discrete event (DE) systems based on a new DE matrix formulation are presented. This new formulation is a hybrid system with logical and algebraic components that allows fast, direct design and reconfiguration of rule-based controllers for manufacturing systems. It applies to general DE systems that include shared resources, dispatching, circular waits, and variable part routing. A certain DE matrix state equation together with the familiar Petri net marking transition equation yield a complete dynamical description of a DE system. Our goal in this article is to show that this provides a simplified computer tool that allows efficient simulation and modeling for DE systems.

Human control strategy: abstraction, verification, and replication

Abstract: In this article, we describe and develop methodologies for modeling and transferring human control strategy. This research has potential application in a variety of areas such as the intelligent vehicle highway system, human-machine interfacing, real-time training, space telerobotics, and agile manufacturing. We specifically address the following issues: (1) how to efficiently model human control strategy through learning cascade neural networks, (2) how to select state inputs in order to generate reliable models, (3) how to validate the computed models through an independent, hidden Markov model-based procedure, and (4) how to effectively transfer human control strategy. We have implemented this approach experimentally in the real-time control of a human driving simulator, and are working to transfer these methodologies for the control of an autonomous vehicle and a mobile robot. In providing a framework for abstracting computational models of human skill, we expect to facilitate analysis of human control, the development of human-like intelligent machines, improved human-robot coordination, and the transfer of skill from one human to another.

Fuzzy guidance controller for an autonomous boat

Abstract: A fuzzy guidance controller is developed for waypoint following guidance on a small autonomous boat. The boat is a prototype vehicle developed at the MIT Sea Grant College Program and is designed to perform autonomous oceanographic research. The fuzzy controller is programmed in C and down-loaded to a small onboard computer for execution. Navigation data is found in real-time using a digital compass and a differentially corrected GPS receiver. The fuzzy controller was found to be relatively easy to develop, simple to tune, and robust to external disturbances. Field test results are presented and show that the guidance controller performed well on complex paths.

Implementation of a friction estimation and compensation technique

Abstract: A friction estimation and compensation technique was implemented on a laboratory apparatus designed to permit the direct measurement of friction. Experimental results are reported for a friction observer which estimates total friction present assuming it to be a constant times the sign of velocity. A second observer is used to estimate the velocity using the measured position of the rotating shaft in the apparatus, when velocity is not measurable. Experimental results show that the friction estimate is consistent with the measured friction, displaying the theoretical hysteresis phenomenon. Moreover, the performance of the system is substantially improved by the use of the estimated friction to compensate the system, especially at very low velocity.

Position-based impedance control of an industrial hydraulic manipulator

Abstract: This article addresses the problem of impedance control in hydraulic manipulators. Whereas most impedance and hybrid force/position control formulations have focused on electrically driven robots with controllable actuator torques, torque control of hydraulic actuators is a difficult task. A position-based impedance controller (PBIC) is proposed and demonstrated on an existing industrial hydraulic robot (a Unimate MKII-2000). A nonlinear proportional-integral (NPI) controller is first developed to meet the accurate positioning requirements of this impedance control formulation. The NPI controller is shown to make the manipulator match a range of second-order target impedances. Various experiments in free space and in environmental contact, including a simple impedance modulation experiment, demonstrate the feasibility and the promise of the technique. Finally, explanation of an experimentally observed behaviour is offered, suggesting a basic limitation to the implementation of impedance control.

Identifying chaotic systems via a Wiener-type cascade model

Abstract: In this article we first show a theory that the Wiener-type cascade dynamical model, in which a simple linear plant is used as the dynamic subsystem and a three-layer feedforward artificial neural network is employed as the nonlinear static subsystem, can uniformly approximate a continuous trajectory of a general nonlinear dynamical system with arbitrarily high precision on a compact time domain. We then report some successful simulation results, by training the neural network using a model-reference adaptive control method, for identification of continuous-time and discrete-time chaotic systems, including the typical Duffing, Henon, and Lozi systems. This Wiener-type cascade structure is believed to have great potential for chaotic dynamics identification, control and synchronization.

Variable damping impedance control of a bilateral telerobotic system

Abstract: Damping and stiffness control in a telerobotic system allows the programmer to define the master and slave dynamics to suit a given task. Unfortunately, many requirements for good teleoperation produce conflicting choices of desired dynamic parameters for different tasks. It is proposed here that the master and slave damping and stiffness matrices be functionally dependent on sensed and commanded values of force and velocity, with no previous knowledge of the environment required. A strategy has been devised which provides better quality teleoperation under a variety of circumstances than is achievable with constant dynamics. Tracking in free space and along a surface by the slave can be maintained while impact forces are minimized with this strategy. The variable damping algorithm has been implemented on a 7-degree-of-freedom Robotics Research Corporation manipulator with position input from a 6-degree-of-freedom Kraft master hand controller, and tracking and impact performance is illustrated.

Complex adaptive systems: concepts and power industry applications

Abstract: Research in complex adaptive systems (CAS) has begun to achieve an understanding of complexity in natural systems as a phenomenon that emerges from the interaction of multiple, simple, but adaptive, agents. The computer experiments, or simulations, used in this research have led to the development of mathematical and computational techniques that are equally applicable to the design of distributed control systems based on the model of a complex system composed of multiple, autonomous, intelligent agents, competing and cooperating in the context of the whole system's environment. The electric power grid, made up of many, geographically dispersed components, is itself a CAS that can exhibit global change almost instantaneously as a result of local actions. The new availability of very high voltage active control devices makes the grid a prime candidate to benefit from a distributed control system based on the CAS model. Such a control system for the grid may be not only useful, but necessary, in order to make feasible the current worldwide trend toward free competition in electric power, which, carried to its ultimate extreme, requires the substitution of coordination by individual agents in a free market for centralized control from a single command site. Using autonomous, adaptive agents to model and simulate the corporate entities, involved in this free competition, can also evolve new business strategies for internal reorganization, external partnerships, and market penetration.

Algorithms and software for LMI problems in control

Abstract: A number of important problems from system and control theory can be numerically solved by reformulating them as convex optimization problems with linear matrix inequality (LMI) constraints. While numerous articles have appeared cataloging applications of LMIs to control system analysis and design, there have been few publications in the control literature describing the numerical solution of these optimization problems. The purpose of this article is to provide an overview of the state of the art of numerical algorithms for LMI problems, and of the available software.

Neurofuzzy model-based weld fusion state estimation

Abstract: Proper fusion is crucial in generating a sound weld. Successful control of the fusion state requires accurate measurements of both the topside and back-side bead widths. A top-side sensor-based system is preferred so that the sensor can be attached to and moved with the torch. Thus, the system must be capable of estimating the back-side bead width with high accuracy. Because skilled human operators can estimate the fusion state from the observed weld pool, a neurofuzzy system is developed to infer the backside bead width from the pool geometry in this work. It is found that the back-side bead width can be estimated with satisfactory accuracy by the identified neurofuzzy model. Thus, accurate feedback of the fusion state can be provided for its control.

Robust detection and isolation of process faults using neural networks

Abstract: The problem of robust model-based diagnosis of process faults is addressed by means of artificial neural networks. Different structures and learning methods are investigated for both approaches to function approximation and pattern recognition. Main emphasis is placed upon static and dynamic neural nets that are used as predictors of nonlinear models for symptom generation. Dynamic neural networks are properly integrated into a generalized observer scheme. The goal is to achieve an adequate approximation of process outputs for each known class of system behavior. Symptoms are then evaluated by means of pattern classification. Application to a laboratory process is presented. A diagnosing subsystem is designed to detect incipient faults in the components of a three-tank system. It is implemented in real-time by using the SIMULINL/MATLAB programming environment. Experimental results regarding the diagnosis of single and multiple faults are included in a comparative study. It demonstrates the effectiveness of the suggested approach.

Multiobjective H/sub 2//H/sub /spl infin// guaranteed cost PID design

Abstract: This article presents a methodology for sub-optimal design of PID compensators for systems subject to disturbance signals and to parametric uncertainties of polytope type. The adopted optimality criteria are the H/sub 2/ and H/sub /spl infin// norms of the transfer matrices from the disturbance inputs and from the reference input to the controlled output error. Time constant constraints are also employed in the optimization procedure. The PID parameter selection combines the different optimization criteria through a multiobjective technique. True guaranteed cost values for optimization criteria are calculated. An example is presented, showing the uncertainty polytope construction from physical parameters tolerances and the PID synthesis procedure. A genetic algorithm and extensive simulations are employed in order to evaluate the proposed algorithm performance.

Control using chaos

Abstract: A nonlinear dynamical system with a chaotic attractor will produce motion on the attractor which has random-like properties. This observation leads to a very simple algorithm for bringing a discrete or continuous nonlinear dynamical system to a fixed point. Suppose that the system to be controlled is either naturally chaotic or that chaotic motion can be produced by means of open-loop control. Suppose also that a neighborhood of the desired fixed point can be found, such that, under state variable feedback control, the system is guaranteed to be driven to the fixed point. If this neighborhood has points in common with a chaotic attractor, it may be used as a "controllable target" for the fixed point. The control algorithm consists of first using, if necessary, open-loop control to generate chaotic motion and then waiting for the system to move into the controllable target. At such a time the open-loop control is turned off and the appropriate closed-loop control applied. A basic requirement with this approach is to determine a large enough controllable target so that one does not have to wait too long for the system to reach it. The following method is used here: the system is first linearized about the desired fixed-point solution. If necessary, a feedback controller is then designed so that this reference solution has suitable stability properties. Then a Lyapunov function is obtained based on this stable linear system and a level curve for the Lyapunov function is determined, such that, whenever the state of the nonlinear system is within this level curve, the feedback controller will drive the nonlinear system to the desired equilibrium solution. Such a level curve defines a controllable target provided that it actually does contain points on the chaotic attractor. A multiple step approach for determining the Lyapunov level curve is presented which helps in finding large controllable targets for discrete systems.

Benchmarks for the numerical solution of algebraic Riccati equations

Abstract: Two collections of benchmark examples are presented for the numerical solution of continuous-time and discrete-time algebraic Riccati equations. These collections may serve for testing purposes in the construction of new numerical methods, but may also be used as a reference set for the comparison of methods.

A multivariable neural controller for automatic ship berthing

Abstract: This article describes the development and application of a multivariable neural controller for automatic ship berthing. Following a brief review of various methods employed in automatic ship control, an online trained, backpropagation-based neural network controller is presented. The principal intention is to take advantage of the learning ability of neural networks, and to derive an autonomous neural control algorithm which is independent of the mathematical model of the ship. The proposed neural network controller is designed to adjust its parameters online from a direct evaluation of performance accuracy, thereby eliminating the need for off-line training and a "trainer" associated with supervised control. In addition, the nonlinearity of the rudder and the transfer lag of the propeller have been considered in the system design to increase the realism of the simulation. A series of simulation studies, which include wind disturbances and shallow water effects, have been undertaken to demonstrate the adaptive features and the robust performance of the proposed neural control scheme.

Adaptation for active noise control

Abstract: In active noise control, adaptive approaches play an essential role when dealing with uncertainties and changes in noise and control path dynamics of sound propagation. This article is concerned with two new adaptive schemes. One is a direct adaptive approach which can explicitly update a feedforward controller with guaranteed stability on the assumption that the control path dynamics are known a priori. The other is an indirect adaptive approach which is available when the both noise and control path dynamics are uncertain and changeable. The algorithm consists of online identification of two relevant transfer function models and real-time calculation of the corresponding feedforward controller. The two adaptive schemes are examined in experimental studies in an air duct system and their effectiveness is shown by the results.

Motion planning of a pneumatic robot using a neural network

Abstract: Integration of sensing and motion planning plays a crucial role in autonomous robot operation. We present a framework for sensor-based robot motion planning that uses learning to handle arbitrarily configured sensors and robots. The theoretical basis of this approach is the concept of the perceptual control manifold that extends the notion of the robot configuration space to include sensor space. To overcome modeling uncertainty, the topology-representing-network algorithm is employed to learn a representation of the perceptual control manifold. By exploiting the topology-preserving features of the neural network, a diffusion-based path planning strategy leads to flexible obstacle avoidance. The practical feasibility of the approach is demonstrated on a pneumatically driven robot arm (SoftArm) using visual sensing.

Generalized predictive control for active flutter suppression

Abstract: This article presents experimental results of a transonic wind-tunnel test that demonstrates the use of generalized predictive control for flutter suppression for a subsonic wind-tunnel wing model. The generalized predictive control algorithm is based on the minimization of a suitable cost function over finite costing and control horizons. The cost function minimizes not only the sum of the mean square output of the plant predictions, but also the weighted square rate of change of the control input with its input constraints. An additional term was added to the cost function to compensate for dynamics of the wing model that cause it to be invariant to low input frequencies. This characteristic results in a control surface that drifts within the specified input constraints. The augmentation to the cost function that penalizes this low frequency drift is derived and demonstrated. The initial validation of the controller uses a linear plant predictor model for the computation of the control inputs. Simulation results of the closed-loop system that were used to determine nominal ranges for the tuning parameters are presented. The generalized predictive controller based on the linear predictor model successfully suppressed the flutter for all testable Mach numbers and dynamic pressures in the transonic region in both simulation and wind-tunnel testing. The results confirm that the generalized predictive controller is robust to modeling errors.

A distributed control system for flexible materials handling

Abstract: In this research, we are developing a materials handling system where many small, simple actuators cooperate to convey large objects with three degrees of freedom in a plane. The actuators, or cells, are arranged in a regular array which is fixed to a planar surface, and objects are passed over the array. Such an array provides very flexible materials handling in which many objects can be conveyed simultaneously in different directions. The array is coordinated in a distributed manner, rather than by a central controller. Each manipulator has its own controller, and each controller communicates with its neighbours. In this article, modeling and control methods, and a real-time communication network and language are developed. Simple tests on hardware and in simulation are presented.

Teleoperation based on the adjoint Jacobian approach

Abstract: In this article we introduce a method for telecontrol of nonredundant slave arms based on the adjoint Jacobian approach. Cartesian velocity command inputs are used to teleoperate the slave arm smoothly within the whole workspace, including the vicinity of singularities. Moreover, teleoperation at a codimension one singularity is also possible, yielding specific motion patterns. Those include the motion through the singularity, resulting in a reconfigured kinematic chain. The method presented here is, in fact, an alternative to the singularity-consistent null space based approach developed earlier. We discuss some implementation details and provide simulation and experimental data derived during teleoperation of a virtual planar 2R arm and the wrist of a real robot.

Cell mapping for controller design and evaluation

Abstract: Cell mapping is a powerful computational technique for analyzing the global behaviour of nonlinear dynamical systems. It simplifies the task of analyzing a continuous phase space by partitioning it into a finite number of disjoint cells and approximating system trajectories as cell transitions. The resulting cell map provides global measures of stability and other performance characteristics that are valuable in system analysis and controller design. This article shows how cell mapping can be used to design high-performance, conventional and fuzzy, controllers. It also shows how cell maps can provide global performance measures of the designed controllers, including time optimality, controllability, and empirical assessments of robustness. Evaluating controller performance based on these global measures is superior to simply examining time domain responses for various initial conditions.