Performance evaluation is critical for ensuring the accuracy, efficiency, and reliability of industrial robotic arms. Traditional measurement methods, including contact-based techniques (e.g., coordinate measuring machines and ball-bar systems) and non-contact systems (e.g., laser trackers and optical coordinate measuring machines), offer high precision but are often costly, complex to install, and constrained by environmental factors. To address these limitations, this study proposes a SLAM-based performance evaluation method that leverages LiDAR to track robotic motion without requiring external calibration references. This approach provides a cost-effective and flexible alternative to conventional metrology techniques. However, integrating SLAM into the ISO 9283 framework presents challenges related to accuracy, stability, and measurement consistency. To assess its feasibility, this study evaluates the SLAM-based system by analyzing key performance parameters, ensuring its alignment with industrial requirements. The results demonstrate that the LiDAR-based SLAM system achieves an RMSE of 0.0353 mm in trajectory estimation, confirming its precision and stability. These findings validate the system’s capability as a reliable benchmarking tool for robotic arm performance assessment.

This study develops a dual-arm robot teleoperation system via virtual reality. The objective is to make user control the dual-arm robot through hands and robot vision via HoloLens2. The gestures of the user’s arm and hand are captured and rendered in Unity environment. To map the robot movement with the user’s hand pose, incremental motion mapping and orientation joint control are proposed, and the joint angles are solved through damped least square inverse kinematics. The robotic hands are also designed as end-effectors for multiple tasks. The simulation and experimental results show that the motion of end-effectors matches the user’s hand trajectory, which is synchronized with the user’s movement. The user also completes tasks in several scenarios with the utilities of robot teleoperation. The system constructed in this study combines the actions of humans and robots, promoting multiple robotics research in the fields of imitation learning and remote control.

Robots can deal with different kinds of challenges using tactile sensing arrays as a primary resource. This paper demonstrates the ability of robotic hands to recognize objects' shapes using only a flexible tactile sensor arrays attached to the robotic hand's surface without building the 3D models of objects. A telemanipulation module was developed to achieve a co-moving mechanism between the robotic hand and human hands so that the robotic hand can directly learn the best way to grasp objects from human hands without additional path planning process. Tactile array data were collected while the robotic hand was performing a reiterative grasping process. By extracting the proper features from the tactile sensor array data, the support vector machines (SVMs) were employed to perform object classification. From experiments, the proposed method can achieve 96.67% classification accuracy based on sensory data and SVMs.

We study the modeling of Lagrangian systems with multiple degrees of freedom. Based on system dynamics, canonical parametric models require ad hoc derivations and sometimes simplification for a computable solution; on the other hand, due to the lack of prior knowledge in the system's structure, modern nonparametric models in machine learning face the curse of dimensionality, especially in learning large systems. In this paper, we bridge this gap by unifying the theories of Lagrangian systems and vector-valued reproducing kernel Hilbert space. We reformulate Lagrangian systems with kernels that embed the governing Euler-Lagrange equation-the Lagrangian kernels-and show that these kernels span a subspace capturing the Lagrangian's projection as inverse dynamics. By such property, our model uses only inputs and outputs as in machine learning and inherits the structured form as in system dynamics, thereby removing the need for the mundane derivations for new systems as well as the generalization problem in learning from scratches. In effect, it learns the system's Lagrangian, a simpler task than directly learning the dynamics. To demonstrate, we applied the proposed kernel to identify the robot inverse dynamics in simulations and experiments. Our results present a competitive novel approach to identifying Lagrangian systems, despite using only inputs and outputs.

We investigate the modeling of inverse dynamics without prior kinematic information for holonomic rigid-body robots. Despite success in compensating robot dynamics and friction, general inverse dynamics models are nontrivial. Rigid-body models are restrictive or inefficient; learning-based models are generalizable yet require large training data. The structured kernels address the dilemma by embedding the robot dynamics in reproducing kernel Hilbert space. The proposed kernels autonomously converge to rigid-body models but require fewer samples; with a semi-parametric framework that incorporates additional parametric basis for friction, the structured kernels can efficiently model general rigid-body robots. We tested the proposed scheme in simulations and experiments; the models that consider the structure of function space are more accurate.

Human intention estimation is important for assistive lower limb exoskeleton, and the task is realized mostly by the dynamics model or the EMG model. Although the dynamics model offers better estimation, it fails when unmodeled disturbances come into the system, such as the ground reaction force. In contrast, the EMG model is non-stationary, and therefore the offline calibrated EMG model is not satisfactory for long-time operation. In this paper, we propose the self-learning scheme with the sliding mode admittance control to overcome the deficiency. In the swing phase, the dynamics model is used to estimate the intention while teaching the EMG model; in the consecutive swing phase, the taught EMG model is used alternatively. In consequence, the self-learning control scheme provides better estimations during the whole operation. In addition, the admittance interface and the sliding mode controller ensure robust performance. The control scheme is justified by the knee orthosis with the backdrivable spring torsion actuator, and the experimental results are prominent.

Human behaviors are highly affected by implicit human factors such as culture, social conventions, laws and even the mental states of individuals and groups. If robots are to be accepted by humans, they must conform to common social norms and local customs and recognize highly socialized spatial behaviors. This paper aims to extend the Spatial Behavior Cognition Model (SBCM) proposed in [14] by adding the pedestrian velocity as one of the estimated features of SBCM. In this case, robots can understand more spatial behaviors represented mainly on pedestrian velocity. Experiments for several scenarios of human environments are conducted. The results show that the proposed approach enables robots to accumulate the knowledge needed to ensure good behavior in social situations.

In the near future, robots will exist in our home, office, hospital, school and all other public places. In other words, robots have to work and share environments with human. However, there exist many invisible social rules or social protocols in human society. For example, people usually stand in a line to access certain services. It seems that there is a social force or a “spatial effect” which forces people to perform the behavior of standing in a line. Unfortunately, these spatial effects are usually invisible and immeasurable. It makes social behavior understanding to be a difficult problem. The robot that does not comprehend these spatial effects might harm people or itself. Previous methods of behavior understanding mainly only targeted certain scenarios or environments. This paper follows the framework of SBCM (Spatial Behavior Cognition Model)[3] and extends to feature-based spatial effects. By utilizing feature to represent the spatial effect, it is easier to apply known spatial effects to different environments. This paper also demonstrates that the robot is able to learn and execute some common social behaviors under the proposed method. Several simulations and experiments are conducted to verify the proposed idea in this paper.

Simultaneously considering the physical interaction between the user and the robot within safety and performance constraints in rehabilitation and human walking situations, this paper proposes a new backdrivable torsion spring actuator (BTSA) with hybrid control that switches between direct electromyography (EMG) biofeedback control and zero impedance control, to provide a novel rehabilitation training and walking assistance mechanism for humans. The proposed backdrivable 1-DOF serial elastic actuator is designed to achieve intrinsic safety, compliance properties, and control performance. The proposed mechanical system can provide desirable backdrivable property and softer stiffness than that of traditional robots. In additional, the proposed hybrid control not only considers the assistive function, when human assistance is required, but also the compliance property, when assistance is not needed. Compared to state-of-the-art assistive methods, the BTSA with the proposed hybrid control system is unique in that it can simultaneously achieve assistance control through EMG biofeedback and compliance control through zero impedance control. A simple human-robot interaction model is built to investigate performance and explain the whole control concept. Further, a knee exoskeleton is built and three kinds of controls are used on a human subject to demonstrate the difference between them. Both simulation and experimental results show that the proposed BTSA mechanism with hybrid control offers the desired properties.

Focusing on the physical interaction between people and machines within safety constraints in versatile situations, this paper proposes a new, efficient actuation approach, continuous-state coupled elastic actuation (CCEA), to provide oncoming human-machine systems with an intrinsic programmable stiffness capacity to shape output force corresponding to the deviation between human motions and set positions of the system. As one of all the possible CCEA systems, a prototype of a 2-DOF coupled elastic actuator is designed to provide a compromise between performance and safety. Using a pair of antagonistic four-bar linkages, the inherent stiffness of the system can be adjusted dynamically. Compared to the state-of-the-art variable stiffness actuators, the CCEA system is unique in that it can achieve near zero mechanical stiffness in an efficient way. In addition, a human-robot interaction model is built to investigate the controlled bandwidth and safety of the CCEA system. For the application of assistive exercises, this study also proposes two kinds of controls for assistive exercises. Finally, a CCEA exoskeleton is built for elbow rehabilitation. Both simulations and experiments are conducted to show some desired properties of the proposed CCEA system.

Many diseases affect human movement functions and daily living. In order to recover the ability of movement, rehabilitation is the only way to improve the situation for many clients. Therefore, this study proposed a novel robot with brain-computer interface for rehabilitation exercises, namely BCRS (Brian-Controlled Rehabilitation System) and using multiple kernel learning (MKL) trains the classifier. BCRS detects and classifies the P300 and non-P300 signals from human brain and determines what kind of the rehabilitation exercises will be chosen. Three types of exercises, passive range of motion, isotonic, and isometric exercise, were realized in the system and support vector machine was used as the classification algorithm. For the three exercises, a new P300 panel was designed and composed of 25 commands. Through the experiments, we can find that BCRS can achieve good performance for rehabilitation exercises and MKL is a good method for EEG to have good accuracy of P300 signal classification and low training time than support vector machine (SVM).

In human society, there are many invisible social rules or spatial effects existing in our environments. The robot that does not comprehend these spatial effects might harm people or itself. This paper presents a spatial behavior cognition model (SBCM) to describe the spatial effects existing between people and people, people and environments. By understanding the spatial effects in human-lived environments, the robot not only predicts pedestrian intentions and trajectories but also behaves socially acceptable motions. Moreover, the concept of pedestrian ego-graph (PEG) is proposed to efficiently query pedestrian-like paths for trajectory prediction. Model evaluation and experiments are shown to verify the proposed idea in this paper.

For classical robotic applications, robotic systems consist of servo motors, high-ratio reduction and rigid links; mechanical designers prefer to designing robotic applications as stiff as possible to make robots manipulate with remarkable speed and precise position movements. However, these robotic applications can hardly interact with people and environments under safety constraints. It poses the very fundamental problem of ensuring safety to humans and protecting the robot. This paper presents an active-passive variable stiffness elastic actuator (APVSEA) which is designed for safety robot systems. The APVSEA consists of two DC-motors: one is used to control the position of the joint and the other is used to adjust the stiffness of the system. The stiffness is generated by two antagonistically nonlinear springs. By changing the preload length of the two antagonistically nonlinear springs, APVSEA has the ability to minimize large impact forces due to shocks, to safely interact with the user and/or become as stiff as possible to make precise position movements or trajectory tracking control easier. Experiment results are presented to show that APVSEA is capable of providing precise position movements while offering safe human-robot interaction.

We present a novel one-class classification method called conformal-kernel support vector data description (CK-SVDD) for P300 speller brain-computer interface (BCI). The CK-SVDD has lower computational complexity than the widely-used SVM, and better classification accuracy than both SVM and SVDD, thus being able to improve the usability of the P300 speller when it is used in an online mode. We also developed a real-time EEG acquisition and preprocessing module, as well as the software written in C#, which executes the functions of data processing and classification. The results, carried out on three subjects, show that the proposed CK-SVDD consistently performs better than the SVM and the original SVDD in different numbers of rounds. The results also indicate that it can achieve a high character classification accuracy of over 95% for all subjects, even when the number of rounds is less than 4 in some cases, thus being able to provide a fast online testing speed without losing classification accuracy.

The main objective of this paper is to develop an autonomous navigation system for robot arms so that they can operate in complex environments, such as kitchen, factory, etc. In order to solve the problem of planning in the dynamic environments, we proposed a CT-RRTs (bi-direction RRTs in configuration time space) planner, which can make the robot arm reach goal successfully in the partially known dynamic environment while remains the greedy heuristics of the RRT-Connect. It is planned in the augmented state-time space of the robot arm configuration and the positions of the moving objects. Various techniques are used to accelerate planning and enhance its safety. We used the new planner to develop the dual-arm planner, and it can be used as the basis to solve higher level problems in motion planning. The experiments show that the proposed method can effectively resolve the dual-arm motion planning problem.

This paper aims at dealing with a critical issue for electromyography (EMG) recognition. The issue is related to the stability of an EMG-based prosthesis control. Traditional EMG recognition systems receive EMG patterns and send them into classifiers directly, which generally results in unstable situations if the classes of some of the input EMG patterns are not included in the training of the classifiers. The EMG patterns whose class labels are not defined in the training phase are called non-target patterns. There should be a filter and this filter should be able to reject all non-target EMG patterns. As such, only target EMG patterns are fed into classifier, thus achieving a high-accuracy EMG classification. To this end, we propose in this paper a one-class classification-based non-target EMG pattern filtering scheme. By introducing a novel one-class classifier, called support vector data description (SVDD), into the filtering scheme, the goal mentioned above can easily be achieved. SVDD is a powerful machine learning technique. It can be built on a single class and find a flexible boundary to enclose the target class by using the so-called kernel trick. In experiments, we will show that if the filtering scheme is not performed, the traditional EMG classification system suffers from unstable situations. Contrarily, the whole classification system will achieve satisfactory and stable performance no matter what the input EMG patterns are target or non-target ones, if the proposed filtering scheme is embedded.

To achieve fully autonomous mobile robot in unknown environment, a mapping and localization technique is required. Recently, in unknown dynamic environment, SLAMMOT (Simultaneous Localization and Mapping and Moving Object Tracking) is also attracted the extensive attention in this area. In this paper, we extended SLAMMOT problem to simultaneous map prediction and moving object trajectory prediction. The robot not only passively collects the data and executes SLAMMOT, but actively predicts the future scene. The recursive Bayesian formulation of SLAMMOT with scene prediction is also derived for real time operation. Preliminary results are also shown and validated the idea in this paper.

With increasing applications of RFID (radio frequency identification) technology, reliability requirements for the deployment of RFID readers have become more critical. For a robust deployment of a RFID network, a systematic solution was proposed in this paper. To simplify the reader deployment problem, instead of the traditional 2D circle-like shape assumption, the ellipse/ellipsoid-like shape was taken to represent the signal range of an RFID antenna in this paper. Further, the performance of RFID was estimated statistically. In addition, a deployment tool was designed for users to obtain a recommended layout. For a given area, solutions of number and placement of RFID readers are computed fast and robustly. To avoid reader collision which may result in low efficiency of a RFID network, two simplified mechanisms were proposed. Finally, the practical implementation and simulations were performed to test the practicability of the deployment tool and further to confirm the robustness and adequacy of the proposed models.

In this paper, the equations of motion of the 3D dynamical model of the caster wheeled mobile robot moving on the horizontal flat surface are derived. During the process of deriving the equations of motion, the robot body and three wheels of the mobile robot system are viewed as four rigid bodies. The three wheels of the mobile robot are subject to sliding and skidding if the friction force at the contact point can not provide enough resistance to prevent the wheels from slipping. Therefore, the constraint arising from the non-slip condition between the wheels and the ground is released. The Lagrange equations of motion of the system with two inputs are obtained. These derived equations of motion can be expectantly used in many aspects, such as path tracking, adaptive control, and motion control of the mobile robot

In this paper, a new algorithm, relative-absolute map filter (RAMF), is proposed to solve the simultaneous localization and mapping problem. Compared with FastSLAM, which adopts many absolute maps to describe the relationship between features, RAMF utilizes only one relative map instead. By fusing the information of relative map and absolute map, RAMF can create a more accurate map. Moreover, the embedded particle filter in RAMF can handle robot localization. Simulation results show that RAMF has better performance than FastSLAM and UKF SLAM in the noisy robot motion

Production scheduling is a complex task in most real manufacturing settings and semiconductor manufacturing is no exception. To control such complex systems, it is a challenge to determine appropriate dispatching strategies under various system conditions. Therefore, dispatching strategies are important to improve the system performance, especially for the real time control of the system. In this paper, an on-line rescheduling mechanism combined with theory of constraints (TOC) is proposed. Genetic algorithm (GA) is for searching dispatching rule sets which promote better performance. With system conditions corresponding to dispatching rules, the support vector machine (SVM) classifier is constructed as the scheduler. Also, the adaptive neuro-fuzzy inference system (ANFIS) prediction model is built for the sake of on-line deciding the scheduling intervals. The results indicate that applying the proposed mechanism to obtaining dispatching strategies is an effective method dealing the product mix problem and considering the complexity and variation of semiconductor wafer fabrication systems.

This paper develops a wearable biomedical health-care system which incorporates the wearable computing technology with multi-agent software architecture to provide a solution to the medical-care problems. The embedded Linux based wearable medical computer connects various sensors which can detect the patients' vital signs, such as electrocardiogram (ECG), body temperature, and body humidity. In order to get more reliable and accurate information, the multi-sensor fusion algorithm is employed to integrate the sensor information. The filter banks based QRS complex and heart beat detection algorithm, and fuzzy fusion algorithms are implemented to construct the diagnosis system of this wearable biomedical computer. Furthermore, the personal nursing agent (PNA) and multi-agent based infrastructure for modern patient information and management system are developed. The experimental results show that this system is helpful and useful for managing and monitoring patients. Accordingly, the wearable biomedical health-care system is regarded as a highly efficient biomedical system and inherits both features of multi-agent architecture and wearable computing, i.e., decentralized, reliable, autonomous, unrestrictive, and cooperative features.

In this paper, we propose a fast dynamic visibility graph (DVG) for constructing a reduced roadmap among convex polygonal obstacles. DVG is extracted from the global environment with the simple geometric method and rules. Moreover, the data preprocessing is based on the concept of V-circle. Through V-circle, the process is speeded up greatly. Finally, DVG is extended to deal with multi-target problems that traditionally require a lot of time for reconstructing configuration space (C-space).

In this paper, a microscopic 3D image is reconstructed for a micro-manipulation system. The 3D image reconstruction is based on a binocular stereo vision system. We propose a new method called "support value" to solve correspondence problems. Geometric constraints are added to reduce the ambiguities in stereo matching. Left/right consistency check is used to increase the confidence for matching results. Post-processing helps remove the outliers and make original dense disparity map smooth and reliable. The three-dimensional characteristics can be reconstructed by perspective projection. Finally, the 3D image reconstruction algorithm is successfully applied to a micro manipulation system for cell image reconstruction by a simpler and more friendly way.

Electromyograph (EMG) features have the properties of large variations and nonstationary issue in the classification of EMG is the classifier design. The major goal of this paper is to develop a classifier for the classification of eight kinds of prehensile postures to achieve high classification rate and reduce the online learning time. The cascaded architecture of neural networks with feature map (CANFM) is proposed to achieve the goal. The CANFM is composed of two kinds of neural networks: an unsupervised Kohonen's self-organizing map (SOM), and a supervised multi-layer feedforward neural network. Experimental results show that by extracting EMG features, forth-order autoregressive model (ARM) and histogram of EMG signals (IEMG), as inputs, the proposed CANFM can obtain and remain high classification rates compared with other classifiers, including k-nearest neighbor method (K-NN), fuzzy K-NN algorithm, and back-propagation neural network (BPNN) in several online testing.

Electromyograph (EMG) has the properties of large variations and nonstationarity. There are two issues in the classification of EMG signals. One is the feature selection, and the other is the classifier design. Subject to the first issue, we propose a supervised feature mining (SFM) method, which is an intelligent approach based on genetic algorithms (GAs), fuzzy measure, and domain knowledge on pattern recognition. The SFM can find the optimal EMG feature subset automatically and remove the redundant from a large amount of feature candidates without taking trial-and-error. In the experiments, all feature candidates and optimal feature subset are conducted to demonstrate the validity of the proposed SFM. Moreover, experimental results show that the optimal EMG feature subset contained from SFM can obtain higher classification rates compared with using all feature candidates by K-NN method.

Integrated circuit (IC) foundry fabs are difficult to model due to the increasing product mixes and flexible routes. An imprecise fab model cannot be used to evaluate fab performance or estimate product cycle time. The distributed colored timed Petri net (DCTPN) is a type of high-level Petri net with embedded entity attribute, time, communication, and remote object invocation properties that contribute to realistic descriptions, distributed modeling, and manufacturing execution system integration. Furthermore, the resource-oriented DCTPN modeling extends product mix and flexible route modeling capabilities. The DCTPN conflict resolution and token competition rules are used to construct the dispatching system for a controlled IC fab. A simplified 200 mm IC fab with six functional areas is discussed and demonstrated based on different control policies.

A modeling methodology for conceptual design of an automated material handling system using a distributed colored timed Petri net (DCTPN) is developed. The IC fabrication process-flow model is constructed and integrated with AMHS models. In IC fabrication process-flow modeling, a method called virtual tool group is proposed to solve the problem in modeling tool-coupling effects. From this entire fab model, several performance indexes, such as WIP, bottleneck detection, cycle time, wafer operation history tracking, and vehicle utilization can be evaluated from the DCTPN simulation result.

Feature selection is a very important process in a pattern recognition system. In the paper, a GA-based fuzzy feature evaluation algorithm (GAFEA) is proposed. The relative importance of a feature element in the feature space is explored in a 2-D example, and a weighting factor is assigned to each feature element to magnify or reduce its importance in the feature space. In addition, a fuzzy feature evaluation index (FFEI) is defined for the measurement of ambiguity of intraclass and interclass. By integrating the weighting factors with the index FFEI, the index FFEI becomes a function of the weighting set. Based on minimizing FFEI, the degrees of the relative importance of features in the multi-dimensional feature space can be found by the genetic algorithm. Finally, an experiment is conducted to justify the validity of the proposed GAFEA. The results show that the task of the feature evaluation can be achieved by the proposed GAFEA before selecting the optimal subset of features.

A single wafer processing IC fabrication with flexible and distributed configuration is becoming more important. The cluster tool is such an IC fabrication equipment. It consists of a cassette module, transport module, and process module. In the paper, the model of a cluster tool is constructed using a distributed colored timed Petri net. The emulator and controller of the cluster tool are further constructed. The standards of SECS, HSMS and GEM are used to develop the intelligent module controller. Finally, a web-based remote controller is developed and the SPC module is built in the cluster tool.

A virtual unit, tool group, is used to represent the group of tools in semiconductor foundry. Hence, a tool may belong to several tool groups because it may be handle several kinds of recipes. The coupling relation between the tools and the tool groups is a complex many-to-many relation. The paper aims to decide the tool-dispatching rule in tool capacity allocation. The term, "priority-based", is two-fold. One is to prioritize the lots in the fab, the other is to prioritize the tools in the correlated tool groups. The priority-based algorithm is used to estimate the tool utilization, bottleneck, tool-sharing in the tool groups, throughput of the tool groups and WIP (work in process) in a specific day. The proposed algorithm is applied to a real foundry fab and the results show that maximum fuzzy candidative ratio is better than other rules. Besides, a demonstration shows that the supervisor can find the bottleneck tools by the proposed algorithm and release the bottleneck by adding the right tools.

A Chinese musical zither score is different from a western musical staff. The Chinese. zither score is handwritten, and is a combination of fingerings, scales, and several different types of notes. We first construct pattern classes for fingerings and scales we frequently play. A specific segmentation method is derived in accordance with the zither score. After segmentation, all meaningful individuals can be discovered and the weighted cross counting feature is used to extract features. A cascaded architecture of neural network with feature map (CANF) is proposed to obtain high recognition rates. The CANF cascades a supervised neural network trained by back propagation (BPNN) with an unsupervised neural network, Kohonen's self-organized feature map (SOFM). The SOFM can reduce the dimension of feature space and remove the redundancy of features in transformation such that the learning time of BPNN can be sped up and the recognition rate can be improved. In our experiment, a real Chinese zither score is segmented, and the CANF shows a 100% perfect recognition rate.

The keys to the success of a semiconductor foundry fab are based on order fulfilment and customer satisfaction. The main objective of order fulfilment is to deliver the products on time. However, many processes, including due-date setting, planning and scheduling of the order and real-time shop-floor operations, are involved in accomplishing this goal. In order to enhance the customer satisfaction, the foundry company should not only promise on-time delivery but also provide real-time information about orders. A virtual foundry fab can realise these two important issues, and is constructed based on distributed agents and a proposed event-driven platform. The core of the virtual foundry fab is the order fulfilment process. The process is triggered by the customers' orders in the foundry fab environment and handled by the order management system. This paper focuses on the entire order fulfilment process. It includes the order management process, production planning, lot priority setting, tool capacity allocation, shop-floor operation and event monitoring processes. Each subsystem is established as an agent and performs its task on a distributed framework. The results show that the overall virtual foundry fab can be easily constructed using distributed multiple agents.

The electromyographic (EMG) signal is used to discriminate eight hand motions: power grasp, hook grasp, wrist flexion, lateral pinch, flattened hand, centralized grip, three-jaw chuck and cylindrical grasp. From the analysis of the PC-based control system, a three-channel EMG signal is used to distinguish eight hand motions for the short below elbow amputee. Pattern recognition is used in this discriminative system. Three surface electrodes are placed on palmaris longus, entensor digitorum and flexor carpi ulnaris. Due to the complexity of the EMG signal and the portable consideration of the controller, a controller based on digital signal processor (DSP) is designed and implemented in this discriminative system. The DSP integrates the signal preprocessing module, the digital filter module and pattern recognition module into the controller. The online DSP controller can provide 87.5% correct rate for the discrimination of eight hand motions.

A hybrid decomposed queueing network model is developed based on actual operating data from one particular foundry fabrication (fab) for rapid analysis of several performance measures. The queueing model includes analyzer and predictor modules. The system analyzer module provides the analysis of arrival pattern and service pattern for each tool group. The system predictor module provides the forecast of important performance measures, such as products cycle time, lots remaining cycle time, tool utilization, queueing length, tool moves, stage moves. In addition, a modularized system, QFAB, is designed to implement the model proposed. A systematic analysis of arrival pattern and service pattern for tool groups is proposed. An approach to compute the effective tool number for tool groups is also addressed. Based on the analysis, the supervisors can gain more insights and choose the proper queueing models. Comparing the obtained results to the actual fab data, the accuracy of prediction of cycle time is satisfactory.

Proposes the modeling and performance evaluation of a controlled IC foundry fab using a distributed colored timed Petri net (DCTPN). A DCTPN is designed for highly model-mixed and flexible routing manufacturing systems. The distributed models can be integrated through communication places, and the manufacturing execution system (MES) can be executed using a COM (component object model) server place. Especially the process activities of a general IC foundry tool are analyzed, modeled and grouped into a basic tool model library. Based on the tool model library, an entire fab model can be constructed. The DCTPN-based fab model can act as a virtual fab, and be used to estimate fab performance and fab behaviors. In addition, the DCTPN conflict resolution and token competition rules are used to control a fab. Finally, a simplified 200 mm IC fab is presented and its system performance, including throughput, stage move, WIP (wafer in process) distribution, lot cycle time, and utilization, is evaluated. The operation histories of tools and lots are also included. The entire fab model has been verified in a real IC foundry fab.

The performance of a flexible manufacturing system (FMS) depends on the equipment efficiency and process control. In order to increase the equipment efficiency, the failure mode analysis and fault diagnosis can be used to reduce the frequency of unexpected breakdowns of machines. In addition, the statistical process control (SPC) can be used for adjusting the process parameters to eliminate process variations. In this paper, a colored timed Petri net (CTPN) is used to model the process behavior of an FMS. In addition, the CTPN-based SPC, fault diagnosis, and failure model and effect analysis are modeled and analyzed individually. Especially, all of the modular models are integrated and linked based on the CTPN. Due to the unified CTPN modeling, the information of each modular model in the entire system can be exchanged and integrated directly and efficiently. Finally, the entire CTPN FMS models are implemented using G2 real-time expert system. Consequently, the proposed CTPN-based simulator can be acted as a real-time FMS monitor and controller through the G2 standard interface.

A multiple degrees of freedom prosthetic hand, NTU Hand, was developed in our laboratory. The paper focuses on the development of a myoelectric discrimination system for a multi-degree prosthetic hand. The discrimination system uses two surface electrodes to acquire the electromyography (EMG) signal from the flexor digitorum superficialis muscle and the extensor pollicis brevis muscle. Since eight types of hand movements, such as three-jaw chuck, lateral hand, hook grasp, power grasp, cylindrical grasp, centralized grip, flattened hand and wrist flexion are often used in daily life, they are used as key movements in the discrimination system. In order to distinguish those hand movements, the techniques of variance, bias zero-crossings, autoregressive model and spectral estimation are employed for preprocessing of features. Then, an error backpropagation neural network is applied to discriminate among the feature sets. Finally, an analysis interface system based on a PC environment is constructed to verify the idea. Besides, a 3-D graphic interface program based on the programmer's hierarchical interactive graphic system (PHIGS) is applied to simulate the prosthetic movement of the NTU hand. The discrimination system can achieve a success rate of 85% for off-line test and of 71% for online test.

The virtual factory can provide a total solution for shortening manufacturing cycle time and achieving customer satisfactions. Hence, many world-class companies dedicate themselves to establish their own virtual factories. However, it is difficult to implement a virtual factory because of the lack of assistant tools. A three-tier architecture based colored timed Petri net emulator is developed in the paper to provide such kind of tool. The three-tier architecture based emulator is easy to maintain and extend. This paper adopts UML and Microsoft COM/DCOM techniques to design and implement the emulator, respectively. The emulator is suitable to model complex systems, especially those with concurrent and asynchronous behaviours. In addition, this emulator provides distributed simulation ability. Each emulator can communicate with each other to achieve the concurrence among subnets. In particular, the proposed emulator can be integrated with the WWW and a voice system. Finally, the emulator is used to construct a flexible manufacturing system, and demonstrate its feasibility for virtual factory applications.

Both five-finger dexterous hand master (NTU master) and dexterous hand slave (NTU hand) have been developed. The NTU master is driven by tendons and the force feedback is applied by remote electric motors via flexible cable. The NTU master is ten degree-of-freedoms, light weight, portable, easy to wear and has large workspace. In this paper, a DSP-based bilateral control between the master and the slave is developed so that the master can transmit the motion to the slave and receive the perceptual information from the slave by the control. A prediction method is proposed to resolve the time delay due to communication.

Decision trees play important roles in many fields such as pattern recognition and classification It is because they have simple, apparent and fast reasoning process. This paper develops an algorithm to generate a learning fuzzy decision tree. This algorithm firstly collects enough training data for generating a practical decision tree. It then uses fuzzy statistics to calculate fuzzy sets for representing the training data in order to save computing memory and increase generation speed. Finally, this algorithm uses a suboptimal criterion to learn a decision tree from the resultant fuzzy sets. The algorithm is applied to a general-purpose tactile force sensing system. This system uses fuzzy logic to interpolate the force data. Then, the proposed algorithm is used to generate the desired decision tree from the tactile data. Based on the decision tree, the objects can be online recognized precisely.

A distributed colored timed Petri net (CTPN) environment is proposed, based on the object-oriented concept, to create, simulate and model the distributed systems in a convenient way. In this environment, the user can construct a CTPN model easily and then simulate the model in a distributed simulator.

The quality consistence and machine utilization of a flexible manufacturing system (FMS) strongly depend on the statistical process control (SPC) and the correct fault diagnosis of equipment. An FMS can be modelled by the colored timed Petri net (CTPN). However, most CTPN models of FMS lack the activities of SPC and fault diagnosis, and they lead to incomplete FMS CTPN models. In this paper, the CTPN-based SPC and fault diagnosis models are proposed to model the FMS SPC and fault diagnosis behaviors. Since the models depend on the measured data from the inspection machines and the sensor data from the devices, the CTPN-based SPC and fault diagnosis models can be incorporated into the production CTPN models to give a complete model of FMS activities. It is shown that the CTPN-based fault diagnosis is easy to model, and the results are straightforward.

A new five-finger robot hand (NTU hand) with seventeen degrees of freedom (DOF) is developed in this paper. In contrast to traditional tendon-driven robot, the NTU hand has an uncoupled configuration such that each finger and joint are all individually driven. While all actuators, mechanical parts and sensors are packed on the hand, the size of NTU hand is almost the same as a human hand. Such compact design makes the hand easily adapt to industrial robot arms and prosthetic applications. According to the mechanical structure of the NTU hand, the direct and inverse kinematics are also developed in this paper.

Basic concepts about the creep behavior of nonholonomic constrained (NC) systems were discussed previously by us (1995). Two fundamental types of creep kinematics were proposed and the hybrid one could be developed. In this work, we extend to the creep dynamics of typical NC systems such as the disk, sleigh and wheel. First, certain reduced models of constrained motion such as ideal, relative, pair, and general-pair models are explored and related. Secondly, by the invariant manifold method of singular perturbation, fundamental rotational and traverse creep dynamics and a hybrid one are calculated in an approximate way. They can be close to the complete system as possible. Therefore, the bridge between reduced and invariant analyses can be made by the model of general-pair creep and can help us understand physical implications behind the approximate solutions. It is proven that advanced vehicle techniques, such as the anti-lock braking system and a special tracking control system, can be realized by the proposed quasi-constrained creeps.

Most studies about nonholonomic systems were developed under the absolute relation of constraint equations. Such ideally constrained systems are known somewhat relaxed in real world, which are claimed as complementarily constrained problems. Systems with viscous frictions and/or flexibilities are typical cases at which the constraints are pseudo-violated rather than violated. These nonideal kinematics are treated in terms of nonholonomic creep behaviors, classified by two fundamental types: rotation creep and traverse creep. Creep coefficients are defined and proposed to illustrate the kinematical relations for such complementarily constrained systems. It can be shown that most nonholonomic vehicles can be described by both creep types, while the trailers attached to a car falls into the traverse type only. Therefore, we conclude the existing algorithms for nonholonomic motion planning and control should be modified as intrinsic creeps take place in nature. The creep notion can also be applied to the symmetry systems of conservation of angular momentum.

A modified computed torque controller has been developed by N.H. McClamroch and D. Wang (1987). If the mathematical model of the robot is exact, the modified computed torque can simultaneously control the robot motion and contact force in an accurate way. However, there may exist uncertainties in the model, such as flexibility of joints and links, joint friction, and an inexact surface model. It is shown that the modified computed torque controller may result in an unstable closed-loop system for the system with uncertainties. This difficulty can be overcome by using a variable structure controller. The controller is robust in that it is insensitive to variations in the plant parameters and to external disturbances of contact force.>

Four types of constrained robot systems are analyzed: the robot in contact with its environment, the robot with a jig band, two robots carrying a common load, and the robot with internal constraints. A unified formulation of the systems is given using singular systems of differential equations. The formulation poses problems in controller design, trajectory planning, and system simulation that require further work.>

A mathematical formulation for the time-optimal contour following problem, defined by a unilaterally constrained manipulator, is presented. The formulation includes a careful development of the manipulator dynamics and of conditions for avoidance of impact between the end effector and the constraint surface. By a priori specification of the paths for the unconstrained motion segments, a parameterization approach is used to simplify the optimal control formulation, so that solution procedures can be identified. The methodology is applied to a simple contour-following problem for a planar Cartesian manipulator.>

A robotic contour following problem, defined by a unilaterally constrained manipulator, is presented. Our approach explicitly takes into account inequality constraints and resulting contact forces as part of the system dynamics. Possible impact at an entry time is also discussed. A phase plane technique is applied to the timeoptimal motion planning problem subject to conditions of impact avoidance. The contact force is incorporated into the optimal planning problem.