Conference papers by Gawthrop for all years

Keywords: Quantitative Feedback Theory; Robust Control

Keywords: bond graphs, robotics

Keywords:

Keywords:

Keywords:

This paper describes the design of a closed-loop nonlinear controller to improve the dynamic response of a single-spool gas turbine engine. The nonlinear controller is obtained by scheduling the gains of multivariable compensators as a function of engine non-dimensional shaft speed. The compensators, whose outputs are fuel flow and nozzle area, are designed using optimal control theory based on a set of linear models generated from a nonlinear engine simulation. Investigations are also made into developing simple algorithms to obtain an analytical expression for the compressor given its characteristic. The detailed process of developing a nonlinear simulation model for the engine is also described. The open-loop fuel controller is studied using the digital simulation.

Keywords: anaesthesia, estimation

Keywords: anaesthesia, estimation

Keywords:

Conventional bond graph theory is predicated on the notion that a bond has a single causal stroke: an effort imposed at one end implies a flow imposed at the other. This notion is implied by components having a known constitutive relationship.

This paper discusses bond graphs with two causal strokes - Bicausal bond-graphs. These can, for example, handle systems with unknown parameters. This paper addresses issues arising from the general idea of inverse systems - physical systems with a non-standard input-output pattern; in particular: dynamic inverses, parameter estimation and state estimation.

The paper discusses the generation of symbolic models of complex systems using hierarchical bond graphs. The uses to which such models can be put include simulation code generation, linearisation, system inversion for actuator sizing and controller design. This methodology is illustrated with reference to three modelling projects: aircraft systems, plastic-onto-wire extrusion and a gravity wave detector

The continuous-time generalised predictive controller is shown to include the exact-linearisation controller as a special case. An alternative (prediction-free) version of GPC is shown to provide one possible extension of exact linearisation to cope with nonlinear systems with unstable dynamics.

Bicausal bond graphs are used to investigate the controllability of systems in terms of properties of the inverse system.

An approach to the estimation of the physical parameters of nonlinear systems modelled by bond graphs is presented and illustrative examples given.

Real-time simulation requires fast stable integration algorithms with a fixed sample interval which can be chosen without loss of stability. A symbolic approach to generating system-dependent integration algorithms is proposed and evaluated

Some of the theoretical properties of predictive-pole-placement control (a form of model-based predictive control) are given a practical interpretation and corresponding design rules suggested.

A resonant filter approach is proposed for direct identification of continuous-time transfer functions from input-output data when the input contains periodic components. The asymptotic properties of the method are analysed; in particular the noise reduction properties are emphasised. Some illustrative simulations are provided.

Real-time simulation requires fast stable integration algorithms with a fixed sample interval which can be chosen without loss of stability. A symbolic approach to generating system-dependent integration algorithms is proposed and evaluated.

A new approach to time-critical optimisation in the context of identification and control is presented. Symbolic algebra is used to provide the equations for computing the sensitivity functions of a system relevant to provide rapid computation of gradient information. It is verified that the resultant gradient-based optimisation is substantially faster than the corresponding non-gradient method.

Plasticating extruders are an important component in the manufacture of cables and their effective control is vital to the successful dimensional control of manufactured cables.

This paper provides a physically-based model derived from the fundamental physics of the process in a form suitable for model-based control. In so doing, the power of the bond graph approach to multi-disciplinary modelling of complex systems is demonstrated.

The relationship between the Bond Graph and the Behavioural approaches to system modelling are examined and some illustrative examples given. Because of the close links between the two approaches, it is suggested that further investigation would prove fruitful.

The concept of a virtual actuator is shown to simplify the physical-domain design of controllers for systems with non-collocated sensors and actuators.

It is shown that the concept of intermittent control, to be found in the physiological literature, provides a useful alternative to both continuous-time and discrete-time approaches to control system design. In particular an intermittent version of predictive pole placement (PPP) is presented. This intermittent predictive pole placement algorithm is illustrated in the context of a non-linear pendulum-on-cart system.

This paper examines two classes of algorithms that estimate a continuous time ARX type of models from discrete data: one is based on infinite impulse response (IIR) filters while the other is based on finite impulse response (FIR) filters. The IIR filters use continuous time state variable filters, and discretisation is performed on the filtered derivatives. In contrast, the FIR filters are in a discrete form with carefully chosen coefficients to approximate the derivatives of the continuous time variables. The strength and weakness of each approach are discussed and demonstrated by a set of simulation examples.

Physically-plausible models are introduced as simple models which, although they do not pretend to be an accurate models of a particular physical systems, are themselves models of physical systems. A physiological and a thermo-mechanical system are used as examples.

The introduction of product quality self-regulation to food-cooking extrusion is an important aspect of process control within food manufacturing industries. In order to design an automatic control system for product quality, a mathematical model of the food extruder is required. As first-principles models are difficult to obtain in this context, a food extruder is a good candidate for applying system identification tools. This paper presents the application of continuous time system identification to such a food cooking extruder. More specifically, the reported application features an automated identification experiment apparatus designed using relay feedback control mechanisms and instrumented through existing real time supervisory system for the extruder. Experimental data from the food extruder are obtained and analysed using our identification approach.

An intermittent approach to model-based predictive control is successfully applied to an experimental mechanical system -- a “Ball and Beam”. This approach is related to human motion control.

The rapid growth of GNU/Linux in recent years has focused attention on free and open source software. MTT (Model Transformation Tools) is, as far as the authors are aware, the only open source project related to bond graphs. This paper surveys MTT in its present form and invites collaboration from the wider bond graph community.

Bond graph modelling has been applied widely and many ideas have been collected from the user community for improvement, particularly associated with federated systems. A project is underway to address these via a prototype toolset with the working title of Next-Generation Transformation Tools (NTT). This paper will summarise the motivations behind this development and describe notational changes that have been found necessary. This should provide useful insights into the problem of large-scale bond graphs.

The virtual actuator approach is explained and illustrated using three experimental systems. Conclusions are drawn about the importance of transfer system design.

The previously developed Predictive Pole Placement (PPP) controller is modified to give enhanced numerical and stability properties by embedding the method in a linear-quadratic formulation to give a linear-quadratic PPP (LQPPP) controller. Input, output and state constraints are considered using an natural quadratic programming (QP) formulation of LQPPP. Illustrative examples are given.

Although structural constraints such as model order and time delay have been incorporated in the continuous time system identification since its origin, the constraints on the estimated model parameters were rarely enforced. This paper proposes a continuous time system identification approach with constraints. It shows that by incorporating physical parameter information known a priori as hard constraints, the traditional parameter estimation schemes are modified to minimize a quadratic cost function with linear inequality constraints. Using the structure of Frequency Sampling Filters as the vehicle, the paper shows that the constraints can be readily imposed on continuous time frequency response estimation and step response estimation. In particular, a priori knowledge in both time-domain and frequency domain is utilized simultaneously as the constraints for the optimal parameter solution. A Monte-Carlo simulation study with 100 noise realization is used to demonstrate the improvement of the estimation results in terms of continuous time frequency response and continuous time step response.

Experiences at Glasgow University in using LEGO Mindstorms for Control Education are described and implementation details given.

A method for the constrained estimation of the physical system parameters of linear systems from measured data is described which uses a two-stage procedure: step response estimation via the frequency-sampled filter approach followed by a non-quadratic bounded optimisation to obtain the physical parameters. The method is illustrated using simulated data and evaluated using experimental data.

The autonomous underwater vehicles (AUVs) have been developed over three decades for potential uses in scientific, commercial, environmental, and military purposes. The improvement of the computer technology has allowed the expansion of control algorithms into untethered AUV's motions. For these reasons, the paper attempts to derive the method to control the dynamic motions of the vehicle. The nonlinear model of the AUVs is established in six degree of freedom and converts into well known state space model. The model predictive control (MPC) algorithm, using orthogonal functions, is developed to intermittent MPC to manipulate rudder and stern angle signals. The new method allows the MPC to work under strick conditions. An active set quadratic programming procedure is also implemented into the MPC to handle the constrained problems of the dynamic systems

A two-stage identification procedure is applied to the SYSID'06 benchmark example.

A piezoelectric tube is shown to have linear, but non-minimum phase dynamics. The main impediment to the actuation of this piezoelectric tube is the presence of a low-frequency resonant mode which causes mechanical vibrations. A physical-model-based control method is extended to non-minimum phase systems in general and successfully applied to damp the resonant mode; leading to a vibration-free actuation of the piezoelectric tube.

There are a number of competing scientific hypotheses about the structure and parameters of the human control system concerned with balance. System identification techniques have potential to distinguish between such competing hypotheses. As a step towards this goal, the data from an initial series of experiments involving balancing an inverted pendulum by a human via a joystick was analysed using a recently-developed two-stage continuous-time identification method.

This paper proposes a continuous-time model predictive control design for disturbance rejection and set-point following of periodic signals. By assuming input disturbance in the form of sinusoid, the periodic frequency is embedded into the design model. Hence, from internal model principle, the steady-state error of the model predictive control system is ensured to be zero for both disturbance rejection and set-point following. Furthermore, with the design framework of model predictive control, hard constraints on the derivative and amplitude of the control signals are imposed as part of the performance specification. Simulation studies have been used to show the efficacy of the design with or without hard constraints.

Early work on modelling the human motion control system showed that only a part of the corresponding motion signals could be described in terms of a deterministic linear continuous-time model of the human control system. It was suggested that the unexplained part, called the remnant, could be modelled by adding a noise signal with a carefully chosen frequency spectrum. Intermittent control provides an alternative description of the human controller which includes a sampling mechanism. This paper suggests that the remnant can be explained by assuming that this sampling mechanism is not uniform; the addition of a noise signal is not required using this assumption. The approach of this paper is to compare the remnant frequency spectrum derived from experimental data with that from equivalent simulated data using each of the two models of the human controller in turn. It is found that both of the simulated models give similar remnant spectra to that of the experimental data. Further work is required to show which of the two models provides the best physiological explanation of remnant.

Keywords: deterministic linear continuous time model;frequency spectrum;human motion control system;human motor remnant;intermittent control;continuous time systems;motion control;predictive control;

Intermittent control uses open-loop control punctuated with feedback at times determined by error-driven events. The open-loop trajectories are based on an underlying closed-loop strategy and are generated by a system-matched hold. The single-loop event-driven intermittent control method is extended to the multi-loop decentralised control situation. This decentralised intermittent controller is based on an underlying continuous-time decentralised design which is suitable for systems with both input and state interactions. This extension is achieved by using local models of the remote interacting subsystems. These models are used for control signal generation and they are only updated with remote information at discrete event-driven sample times thus reducing information flow. The approach is illustrated using a simulation of a five-pool irrigation channel model previously examined in the literature.

Keywords: closed loop systems;control system synthesis;decentralised control;open loop systems;closed-loop strategy;continuous-time decentralised design;control signal generation;decentralised intermittent control;feedback;five-pool irrigation channel model;multiloop decentralised control;open-loop control;open-loop trajectory;single-loop event-driven intermittent control method;Approximation methods;Bismuth;Control systems;Couplings;Integrated circuits;Mathematical model;Observers

A sensorimotor architecture inspired from biological, vertebrate control should (i) explain the interface between high dimensional sensory analysis, low dimensional goals and high dimensional motor mechanisms and (ii) provide both stability and flexibility. Our interest concerns whether single-input-single-output intermittent control (SISO_IC) generalized to multivariable intermittent control (MIC) can meet these requirements.We base MIC on the continuous-time observer-predictorstate-feedback architecture. MIC uses event detection. A system matched hold (SMH), using the underlying continuoustime optimal control design, generates multivariate open-loop control signals between samples of the predicted state. Combined, this serial process provides a single-channel of control with optimised sensor fusion and motor synergies. Quadratic programming provides constrained, optimised equilibrium control design to handle unphysical configurations, redundancy and provides minimum, necessary reduction of open loop instability through optimised joint impedance. In this multivariate form, dimensionality is linked to goals rather than neuromuscular or sensory degrees of freedom. The biological and engineering rationale for intermittent rather than continuous multivariate control, is that the generalised hold sustains open loop predictive control while the open loop interval provides time within the feedback loop for online centralised, state dependent optimisation and selection.

Keywords: biological techniques;feedback;multivariable control systems;open loop systems;optimal control;quadratic programming;sensor fusion;somatosensory phenomena;biological control;continuous time optimal control design;continuous-time observer-predictorstate-feedback architecture;event detection;feedback loop;joint impedance;motor synergies;multivariate open-loop control signal;neuromuscular;open loop instability reduction;quadratic programming;sensor fusion;sensorimotor architecture;sensory analysis;single-input-single-output intermittent control;system matched hold;unstable multivariate system intermittent control;vertebrate control;Bandwidth;Feedback loop;Joints;Microwave integrated circuits;Muscles;Optimal control;Process control

Efficient energy transduction is one driver of evolution; and thus understanding biomolecular energy transduction is crucial to understanding living organisms. As an energy-orientated modelling methodology, bond graphs provide a useful approach to describing and modelling the efficiency of living systems. This paper gives some new results on the efficiency of metabolism based on bond graph models of the key metabolic processes: glycolysis.