Benjamas Panomruttanarug

Using Kalman filter to attenuate noise in learning and repetitive control can easily degrade performance

Repetitive control (RC) and iterative learning control (ILC) can eliminate deterministic tracking errors of a control system in executing a periodic command or a repeating tracking command. In addition they can cancel errors resulting from a periodic disturbance (RC) or a repeated disturbance (ILC). When there is substantial plant and measurement noise it is natural to consider employing a Kalman filter to improve the error signals used by the RC/ILC law, and the performance is analyzed here. Introducing a Kalman filter to RC or ILC can substantially decrease the steady state error due to noise. However, there are several competing issues. First, when the model used in the filter design is inaccurate, deterministic error is introduced in the response that can be more important than the decrease in error variance from random noise. Second, deterministic steady state errors are also introduced when there are unmodeled repeating external disturbances. Use of a Kalman filter actually requires you to know the time history of the disturbance, not just the period. Hence, one should carefully analyze the situation before deciding to use a Kalman filter. And one should examine model free alternatives to the use of a Kalman filter, such as reducing the learning gain. All of these comments also apply when using a Kalman filter running in time steps in the ILC problem. In third, under appropriate conditions, both ILC and RC are capable of reducing the error level in hardware below the error level in ones model of the system. This very desirable property is lost when one introduces Kalman filtering in the time domain for RC and ILC.

Robustification of iterative learning control and repetitive control by averaging

This paper describes a recently developed averaging technique to robustify iterative learning and repetitive controllers. The robustified controllers are found by minimising cost functions that are averaged over either multiple analytical time-domain models or experimental frequency-domain data. The aim is to produce a technique that is simple and general, and can be applied to any iterative learning control (ILC) or repetitive control (RC) design that involves the minimisation of a cost function. Substantial improvement in convergence to zero tracking error in the presence of model uncertainties has been observed for both ILC and RC by this averaging technique.

Design and analysis of current control for shunt active filter based on repetitive control technique using optimization in the frequency domain

This paper proposes the design and analysis of repetitive control technique for tracking the output current of active filter. The Proportional plus Integral (PI) controller performs a feedback control task and the repetitive control law periodically adjusts the input command to the control system to achieve a desired trajectory. In this work, the repetitive control law is designed using optimization in the frequency domain. The method of finding minimum gains for the optimization cost function is also presented in this paper. Simulation results confirm here the performance of the proposed method.

Bidirectional best first based autonomous parallel parking system

One of the major difficulties with an autonomous reverse parallel parking is how to find a trajectory of travel from start to goal without any collision. In this paper, a bidirectional search is suggested to find the shortest pathway from an initial vertex to a goal vertex in a directed graph, implying that the initial and goal states must be known in advance. In each step for searching, there are three possible nodes generated from the configuration space (or C-space), that the algorithm must decide to traverse to based on a best first search algorithm. Giving a start and goal points, the designed path is obtained by simulation in MATLAB. The experimental results show an ability to perform a parking maneuver using the simulated path.

Frequency domain system identification for servo system in HGA testing system using Particle Swarm Optimization

In the disk drive industry, head gimbal assemblies (HGAs) may be individually tested prior installation in a disk drive. While performing a test, a false negative result may be created because of external air turbulence caused by the rotation of the test disk. The objective of this paper is to find a model of the servo system in HGA testing system so that we know how to adjust the command to get rid of the effect of disturbance. The technique so called Transfer Function Determination Code (TFDC) is proposed to model the system in the frequency domain. In addition, Particle Swarm Optimization (PSO) is used to adjust the weights in TFDC algorithm in each frequency. The simulation results show a comparison of different inertia weight adjustments for the models.

Visual feedback arm robot based on iterative learning control using repetitive control optimization

This paper introduces a use of iterative learning control to improve a performance of arm robot in repeatable task. Iterative learning control aims to converge the repeating distance error produced from arm robot by adjusting the control input in the current run to feedback control system based on error observed from the previous run. The feedback signal to the system is experimentally obtained from the two hardware devices: encoder and webcam. Performance of robot using the two types of feedback signal is discussed in detail. Simple learning control law is first applied to the system and a more complicated control law designed from optimization in the similar field is later used.

Designing Learning Control that is Close to Instability for Improved Parameter Identification

Iterative learning control (ILC) uses an iteration in hardware adjusting the input to a system in order to converge to zero tracking error following a desired system output. Convergence is sensitive to model error, and errors that are sufficiently large to cause divergence, produce inputs that particularly excite unmodeled or poorly modeled dynamics, producing experimental data that is focused on what is wrong with the current model. A separate paper studied the overall concept, and specifically addressed issues of model order error. The first purpose of this paper is to develop modified ILC laws that are intentionally non-robust to model errors, as a way to fine tune the use of ILC for identification purposes. And the second purpose is to study the non-robustness with respect to its ability to improve identification of system parameters when the model order is correct. It is demonstrated that in many cases the approach makes the learning particularly sensitive to relatively small parameter errors in the model, but sensitivity is sometimes limited to parameter errors of a specific sign.

Design of a Noncausal FIR Model Inverse as a Compensator in Repetitive Control

A new method of designing compensators for repetitive controllers is presented. The ideal compensator is a filter that is the inverse of the plant, but this is usually unstable, and therefore cannot be used in practice. The approach used here works on a restricted class of transfer functions, and bypasses this difficulty by making a noncausal FIR model of the plant inverse. This model has poles only at the origin, and is therefore stable. Methods are presented to adjust the three parameters of the design for stability and good learning speed, i.e. the repetitive control gain, and the number of causal, and the number of noncausal gains chosen to compose the finite impulse response model. A third order system is studied, which models the closed loop behavior of one link of a commercial robot. One can produce a stable design with a number of gains ranging from 11 to 15 (this gives the number of real time computations for control update), and these numbers are not particularly sensitive to sample rate. Using 18, 20, or 30 gains can produce a quite reasonable plant inverse model that gives fast learning in repetitions at all frequencies.

Space-indexed aircraft trajectory tracking

Future Air Traffic Management (ATM) demands that flights to be more efficient while maintaining safety and improving flight predictability. Two major research groups Single European Sky (SESAR) and Next Generation Air Transportation system (NEXTGEN) are progressively researching on the concept of 4D trajectory management which aims to improve flight efficiency by integrating time into the 3D trajectory. Therefore the implementation of time-space based flight guidance controller will be investigated in this paper. Firstly, the reference tracking error equations with respect to the spatial variable is defined and later transformed into a reference tracking error with respect to time. Using nonlinear dynamic inversion, the control law is established to make the aircraft accurately follow 3D+T desired trajectories. A simulation is performed to explore the feasibility of the proposed guidance control law and it was found that the proposed guidance control law is able to accurately follow a 3D+T reference trajectory at the same time follow the overfly time constraint.

Traffic Management along Air Streams through Space Metering

Via the current performances of aeronautical communication, navigation and surveillance systems, free flight and traffic management through trajectory negotiation have become a reality. However, the adoption of free flight in congested airspace leads to an increase of the number of potential traffic conflicts which are solved by diverting aircraft from their original flight plan, limiting the benefits of free flight. For high density traffic, air corridor concept and time-based flow management have recently been proposed. In the present paper, it is proposed to organize main traffic flows in congested airspace along air streams which are characterized by a three-dimensional (3D) common reference track and lateral lanes with a dynamic slot structure. There aircraft position is processed in a local space indexed axial coordinates system which should ease the management of traffic separation and surveillance. This change results in the need to develop new 3D space indexed guidance modes to perform position tracking, as well as to design and assign standard trajectories to enter into, evolve inside and exit from the air stream while insuring time and space separation between aircraft.