Shiro Masuda

A direct computation of state deadbeat feedback gains

A method for computing a feedback gain that achieves state deadbeat control is given. From systems given in the staircase form, this method derives the deadbeat gain in a numerically reliable way. It is shown that the gain turns out to be LQ optimal for some weightings. >

A Unified Design Method for Model Reference Adaptive Control Systems Using the Coprime Factorization Approach

Abstract A general design scheme for model reference adaptive control systems is proposed. It is based on the most general exact model matching control law, which has a structure of two-parameter compensator coprimely factored in the ring of proper stable rational functions. It is shown that several schemes proposed by other authors are special cases of the general scheme and their structures are classified and explained. A new design method is also derived through specifying the design parameters in the general scheme. The method allows to place arbitrarily the poles of the closed-loop system.

Q-learning algorithm using an adaptive-sized Q-table

Q-learning is one of the successfully established algorithms for the reinforcement learning. It directly estimates optimal Q-values for pairs of states and admissible actions. By using these Q-values, agents can obtain the optimal movement in controlled Markovian domains without using an explicit model of the desired system. However, the algorithm requires a large number of actions of trial and error in the early stages of learning. In this paper, a Q-learning algorithm using the Memory Based Learning (MBL) system is proposed. By using the generalization property of the MBL system, the learning effect for a Q-value can be spread to adjacent Q-values, and thus the number of actions of trial and error can be reduced. Finally, computer simulation results for the control of inverted pendulums are presented to show the effectiveness of the proposed method.

A design method of multivariable model reference adaptive control system using coprime factorization approach

A design method for multivariable model reference adaptive control systems (MRACS) using a coprime factorization approach is proposed. First, a relation between the coprime factorization over RH/sub /spl infin// and the interactor matrix is derived. Using this relation, a method for design of multivariable MRACS over the ring of proper stable rational functions is obtained. Since the proposed method is based on the coprime factorization approach, the control structure in this paper is more simple. It is shown that the control structure of multivariable MRACS by other authors is included in the structure proposed in this paper. A robust MRAC scheme which achieves asymptotic rejection of unmeasurable disturbances is derived also.<<ETX>>

A construction of multivariable MRACS with fixed compensator using coprime factorization approach

A multivariable model reference adaptive control system (MRACS) with a fixed compensator is proposed. First, a new two-degree-of-freedom (2DOF) compensator with disturbance estimator is derived. Using this structure, a multivariable MRACS with fixed compensator is constructed. Since the proposed method is based on the 2 DOF structure, the fixed compensator is chosen independently of specifications for reference commands. The boundedness of all signals in the closed-loop system and the convergence of the output error are proved. A design method of the fixed compensator for MRACS with low sensitivity is also given. Finally, numerical examples are illustrated in order to show the effectiveness of the proposed method.<<ETX>>

Intersample Performance Improvement in Generalized Predictive Control

Abstract The standard GPC is designed so that a discrete-time cost function is optimized. However, rather poor intersample performance may occur because the cost function in the discrete-time form cannot account for intersample behavior. Therefore, in this paper, a GPC design which can optimize the continuous response using a discrete-time control law with a Zero Order Hold (ZOH) is proposed. To this end, the discrete equivalent plant model and cost function are derived and the modified discrete-time GPC is solved. Finally, a numerical example is illustrated to show the effectiveness of the proposed method.

Transient performance improvement in dynamic certainty equivalent adaptive controllers by including a fixed compensator

We propose a modified Morses dynamic certainty equivalent (DyCE) adaptive controllers in the continuous-time, single-input single-output linear time-invariant plant for the purpose of improving the transient performance. In the new scheme the additive feedback loop through a fixed compensator, which means a non-adaptive one, is included. Furthermore a design method for the fixed compensator is also given, and performance analysis for the proposed DyCE adaptive controller is examined in terms of the mean square tracking error criterion and the /spl Lscr//sub /spl infin// tracking error bound. According to the results of the paper the transient performance can be improved arbitrarily by the properly designed fixed compensator. Finally a numerical example is illustrated in order to show the effectiveness of the proposed method.

A Design Method for a Two-Degrees-of-Freedom GPC Based on a State Space Approach

Abstract This paper considers a new GPC controller designed from the viewpoint of the “two-degrees-of-freedom” compensator scheme based on a state space approach. In the proposed method, the gain of the integrator can be designed independently of the nominal transient response to the reference signal. In addition, the integrator comes into play only in the presence of disturbances and/or model-plant mismatch. Hence, the proposed GPC control law provides overall robustness in the closed loop system because the integrator brings about an undesirable phase lag and often may cause undesirable performance degradation in the controlled processes. The effectiveness of the proposed two-degrees-of-freedom GPC is demonstrated by simulation experiments and experimented evaluation on a computer-interfaced pilot plant.

Multivariable MRACS for systems with rectangular transfer matrix using coprime factorization approach

A multivariable model reference adaptive control system (MRACS) design method for a plant with m inputs and p outputs is proposed (m/spl ne/p). Using an interactor matrix the coprime factorization of the plant for (1) m>p case, and (2) m<p case is derived. Further, using this coprime factorization the structure of MRACS is established.

A Design of an Adaptive Servo Control System Estimating the Internal Model

Abstract An adaptive servo control system estimating the unknown reference signal model for a single-input single-output plant is proposed. The design is based on coprime factorization of the plant model over the ring of stable proper rational functions. An adaptive adjusting law is used to identify the unknown parameters of the reference signal model. Stability of the closed-loop system including the proposed adaptive servo system is proved.