Kazushige Koiwai

Design of a human-skill based PID controller using CMACs

Skilled workers decrease in these years. Then it is a problem that skilled techniques and know-how of specific work are not taken over to next generations. Therefore, a controller which is based on a human-skill is needed to design. On the other hand, in recent year, various neural networks (NNs) have been proposed. These technologies of the NNs enable us to deal with the nonlinear systems, and they play an important role in the field of control engineering. Furthermore, a cerebellar model articulation controller (CMAC) has been proposed as one of artificial NNs. The CMAC is highly abstracted mathematically, considering the cerebellar as a controller for the body motion system. The CMAC has the advantage for the short learning time compared with the conventional NNs. In this paper, a design of human-skill based PID controller using CMACs is proposed. According to the proposed method, PID gains are tuned by some CAMCs which have learned human-skill. Furthermore, the experiment in order to illustrate the proposed scheme was performed by using one of pilot-scaled helicopter control models.

Design and experimental evaluation of an intelligent PID controller using CMACs

Skilled workers decrease in these years. Then it is a problem that skilled techniques and know-how of specific work are not taken over to next generations. Therefore, a controller which is based on human-skill is needed to design. On the other hand, in recent year, various neural networks (NNs) have been proposed. These technologies of the NNs enable us to deal with the nonlinear systems, and they play an important role in the field of control engineering. Furtheremore, a cerebellar model articulation controller (CMAC) has been proposed as one of artificial NNs. The CMAC is highly structured mathematically, considering the cerebellar as a controller for the body motion system. The CMAC has teh advantage for the shor learning time compared with the conventional NNs. In this paper, a skill based intelligent PID controller using CMACs is proposed. According to the proposed method, PID gains are tuned by trained CMACs. Those CMACs are learned using a human skill data in an off-line manner. And also they are trained on-line by a reference model. Furthermore, the experiment in order to illustrate the proposed scheme is performed by using one of pilot-scaled helicopter control models.

Data-driven human skill evaluation for excavator operation

The construction field is increasingly adopting automation and manpower-saving technologies. However, automating all types of construction machinery is difficult because some operations require human skill and judgment-that is, professional skills-for realizing optimal work efficiency. However, the lack of competent professionals because of the decreasing and aging population in Japan is an emerging societal problem and is leading to concerns such as deterioration in workplace safety and difficulty in transferring professional skills. This study proposes human skill evaluation and assesses human skills through the application of control engineering. The human operator is analyzed as a nonlinear controller comprised of a proportional-integral-derivative (PID) controller and a PID parameter tuner. The parameters of this database-driven PID controller are tuned using data (i.e., reference signals, input signals, and output signals) obtained from operation. The proposed approach is experimentally demonstrated by applying it to the case of an excavator.

A consideration on responsiveness evaluation for an excavator based on control engineering approach

Industrial equipment, such as bulldozers, excavators, and cranes, requires human operation. Construction machines with better operability are necessary in order to improve productivity. The evaluation of such operational equipment is performed in order to obtain better operability, at the design stage, by sensory evaluation from a specific evaluator. Therefore, the design and evaluation of operational equipment depend on subjective evaluations. However, subjective evaluation requires a skilled evaluator or plenty of sample data. In this study, a quantitative evaluation is considered with regard to the responsiveness of excavator operation. The evaluated system is approximated by a low-order system, such as a first-order lag system with dead time, based on the control engineering approach. The parameters of the estimated system, i.e., dead time and time constant, are considered in the numerical evaluation of excavator operation responsiveness. The system is operated and evaluated by operators; then, the score by sensory evaluation is obtained. The estimated parameters and evaluated score are investigated.

A consideration on quantitative evaluation of machine responsiveness based on control engineering approach

A lot of Industrial equipments need operations. Equipments such as excavators and cranes are evaluated by sensory evaluations of the specific evaluator at the design stage. Designs and evaluations of operated equipments depend on subjective evaluations. However, subjective evaluations require the skill of evaluation or a lot of sample data. In this paper, a quantitative evaluation is considered about the responsiveness for the machine operation. Based on the control engineering approach, the system which needs the operation are approximated by the first order lag system. System parameters of the dead time and the time constant are considered as the numerical responsive evaluation for the operated machine like a excavator.

Design of a hierarchical-clustering CMAC-PID controller

Proportional-integral-derivative (PID) control algorithm is playing an important role in industrial control process. However, for nonlinear control objects, it is difficult to obtain the desired control performance by using a typical PID controller. Therefore, some intelligent PID controllers are proposed and one of them is a PID controller with some cerebellar model articulation controllers (CMACs). In this controller, CMACs are utilized as a control parameter tuner and PID control parameters are calculated as the output of CMAC. The CMAC obtains higher accuracy by increasing the number of label for each weight table while larger memory is needed and generalization ability decreases. On the other hand, the CMAC costs less memory while obtains higher generalization ability and the accuracy decreases. Hence, a novel CMAC that number of label for each weight table can be decided respectively is proposed so that the accuracy is compatible with the generalization ability. Moreover, the efficiency of the memory allocation is improved. The proposed method employs hierarchical clustering to perform specified numbers of label for each weight table. At last, the effectiveness of proposed method is verified by applying to a nonlinear system numerically.

Design of a data-oriented evolutionary controller for a nonlinear system

In order to design a controller for the nonlinear system, the model-based design method that makes a mathematical model for the system has been proposed. However, the requirement for the accuracy of the model to get a desired control performance is high. Therefore, a data-oriented control method was proposed to prevent making mathematic model for the system. Still, in some cases, a desired control performance is not able to be obtained because the structure of the controller has been fixed when determining the controller. In this paper, a new controller design method that determines the controller using a set of closed-loop data is proposed. With this method, both the control structure and the control parameters can be determined simultaneously. The control structure is decided by the genetic algorithm, and the control parameters are computed by using the least squares method. In this way, a controller that can provide a desired control performance is determined. Furthermore, some simulation examples are verified.

Design of a performance-driven PID controller for a nonlinear system

In industrial control processes, it requires fast response for transient state and for steady state it is necessary to maintain user-specified control performance to achieve desired productivity. Moreover, the Proportional-Integral-Derivative(PID) control algorithm is widely used in industrial control processes. Hence, in this paper an algorithm to tune PID control parameters that can improve control performances of transient state and steady state is proposed. In proposed method, the PID control parameters are tuned by using cerebellar model articulation controller(CMAC) for both in transient and steady state. At the same time, in order to avoid the error that may caused by modeling of system, the CMAC is combine with the algorithm of fictitious reference iterative tuning (FRIT). Therefore, the PID control parameters can be off-line tuned without using system parameters. At last, the effectiveness of the proposed method is numerically verified by using a simulation example.

Human Skill Quantification for Excavator Operation using Random Forest

In the construction field, the improvement of the work efficiency is one of important problems. However, the work efficiency using construction equipment depends on their operation skills. Thus, in order to increase the work efficiency, the operation skill is required to be quantitatively evaluated. In this study, the Random Forest, one of machine learning method, is adopted as the quantitatively evaluation for the operation skill of construction equipment. Evaluated target is the operation on an excavation to load onto a truck for a hydraulic excavator. The Random Forest learns to classify some states by the pilot pressure of skilled worker’s operation. States are defined as ‘dig’, ‘lift’, ‘dump’, ‘reposition’, and ‘idle’. The Random Forest with the learning result of skilled worker is applied to other worker’s operation. The human skill quantification is verified based on the ‘idle’ state.

Feature Extraction for Digging Operation of Excavator Based on Data-Driven Skill-Based PID Controller

Improvement of the work efficiency is demanded by aging and reducing of the working population in the construction field, so that some automation technologies are applied to construction equipment, such as bulldozers and excavators. However, not only the automation technologies but also expert skills are necessary to improve the work efficiency. In this paper, the human skill evaluation is proposed by the data-driven skill-based PID controller. The proposed method is applied to the excavator digging operation. As the result, the difference between the novice operation and the skilled operation is extracted. Moreover, the numerical difference is clarified based on the result.