Rockwon Kim

Data integration and arrangement in the shop floor based-on time stamp: An illustrative study of CNC machining

According to the occurrence period of data in the shop floor, it can be classified into two types: the low-frequency data such as the measurement value and the high-frequency data such as the spindle speed. For the statistical processing or analysis required at the manufacturing site preferentially these two types of data, in which each has its own time-stamp, have to be integrated and arranged. In this paper, we introduce a simple data arrangement algorithms through the comparison of time slots between time stamped data sets that have different frequencies. In order to collect data that meets a desired cycle in the system to handle the collection and arrangement of the data at the same time it should be able to monitor the workload of the system. We show the experimental results (or workload) of proposed algorithms for the case of managing time stamped data sets on the relational database management system (RDBMS).

Building a hierarchical robot task from multiple task procedures

To make better use of a service robot, which interacts with users during the task performance in everyday life, developers and/or users may frequently want to change action sequences so that a robot behaves in a natural manner. Emphasizing the procedural aspects of human task knowledge, we propose a method to generate a FSM-based robot task model from task procedures in possible scenarios. Proposed method consists of description of task procedures, conceptual definition of actions and states, and transformation from task procedures to a robot task language based on hierarchical finite-state machine. A simple illustrative example was described to show the application of the method.

Recovery policies on a Web services-based business process for minimizing human intervention

Web services enables application developers to implement a new application by composing several Web Services. As the recovery mechanisms of Web services composition specifications are insufficient for automatic recovery of system faults, we propose recovery policies for the fault handling of Web services-based business processes. The suggested recovery policies enable the business process execution system to automatically recover the business process from a system fault. As being determined at design time, the recovery polices brings about the reduction of recovery time at run time. In addition, the recovery policies prohibit unexpected side effects resulted from wrong recovery behavior determined at run time

Transformation algorithms between WSCI and BPEL4WS for the collaborative business process

Web services is an emerging technology of distributed computing as an interoperable means of integrating loosely coupled Web applications. To support the business process based on Web services, several specifications, such as BPEL4WS, WSCI, and BPML, were suggested. Among them, BPEL4WS and WSCI may be used to describe a collaborative business process that describes visible message exchanges covering one or more participants involved. In this paper, we suggest the transformation algorithms between WSCI and BPEL4WS to enhance the interoperability. For example, a BPEL4WS implementation system may use the transformation algorithm from WSCI into BPEL4WS to refer to a WSCI-formatted abstract business process

A coordination model for agent behaviors using hierarchical finite state machine with inter-level concurrency

The hierarchical concurrent finite state machine (HCFSM) model natively supports concurrency, hierarchical decomposition, and event handling, which are essential for robot task executions. Moreover, HCFSM is a user-friendly diagrammatic representation model. The operational semantics of this model is defined under the two fundamental assumptions: synchrony hypothesis and zero-time assumption. This simple and theoretic assumption is suitable for time critical tasks rather than high-level behavior coordination including some operational latency time by detecting events and determining behaviors. In this paper, we propose an extended model of the hierarchical finite state machine (HFSM) using inter-level concurrency (IC), called HFSM-IC, which is focused on the intuitive description with easy-to-use constructs. IC replaces the concurrency in HCFSM assuming synchrony hypothesis and zero-time assumption with the multi-thread concept and enables each threaded HFSM to have its own iteration cycle as constant as resources are available.

Data Acquisition for Control Level Automation for SMEs: Requirements and Architecture

This paper proposes the architecture for a system to acquire the real-time data on the control level in the shop floor. The main role of proposed system is to generate labeled dataset while tracking a process and extracting its context. The generated dataset is combined the heterogeneous data from the sources such as production equipment, ad-hoc sensors, and key-in devices. In addition, the context is derived from a set (or a sequence) of processes, procedures, operations, or phase instances defined in ISA-88. Finally, the dataset and the context are used for the predictive analysis on the control level of the automation system.

Applications of Machine Learning Algorithms to Predictive Manufacturing: Trends and Application of Tool Wear Compensation Parameter Recommendation

The manufacturing industry has become more competitive because of globalization and fast change in the industry. To survive from the global market, manufacturing enterprises should reduce the product cost and increase the productivity. The most promising way is applying the information communication technology especially machine learning algorithms to the traditional manufacturing system. This paper presents recent trends of applying machine learning techniques to manufacturing system and briefly explains each kind of applications. As a representative application of machine learning algorithms to manufacturing system, a generalized tool wear compensation parameter recommendation framework using regression algorithms and preliminary results using real data gathered from local and small manufacturing are also presented.

Event-Driven Notation and Hierarchical Automata Semantics for Incremental Task Modeling

Abstract Task modeling and specification for autonomous agents (e.g. robots) normally start with exploring typical scenarios and situational variations. Numerous behavioral procedures of an agent are thus collected. Those procedures are partially different from one another, varying in fragmentation or in abstraction levels. To describe task procedures having this characteristic, we propose an Event-driven Task Script (ETS) notation and a formal semantic framework for modeling agents tasks directly from a set of redundant and partially described scripts. Sequential and parallel behavior control and an event handling mechanism are supported by ETS. The semantics of ETS notation is defined by the corresponding Hierarchical Automata (HA), which is concise and widely used in the area of discrete event systems. The semantic mapping from ETS notation to the semantic domain of HA is also formally defined. An HA task model is incrementally constructed from the ETS scripts. This model transformation enables verification of possibly incomplete task scripts in the HA model. As an illustrative example, we provide a virtual task scenario for a museum guide robot and describe the process of the model transformation.

An experience of the OPRoS component based programming through the implementation of DARwIn robot soccer

In this paper, we present the experiences on the OPRoS component based programming by applying OPRoS to a shooting behavior for a small humanoid robot. We also compare OPRoS with the conventional robot programming method. From the experience and comparison, we expect that user will get some help in choosing the programing methods and the way of programming in OPRoS.