AUTOMATION TECHNOLOGY:EMERGING TRENDS
EMERGING TRENDS
Automation technology has reached a new era. An automation system is automated not only to reach the setting point but to situate itself intelligently in its complex environment. The individual auto- mated systems are also networked to accomplish collaborative tasks. However, networking automated systems is not an easy task. It involves the technologies from the physical levels, which handle the interface and message conversion between automation systems, to the application level, which handles mostly the social interactions between the automation systems. Three typical ongoing research di- rections are described next to present the trends of automation technology.
Virtual Machines
Traditionally, a hierarchy structure of the computer control system for a fully automated manufac- turing system can be divided into four levels (Figure 10). The supervisory control is responsible for managing the direct digit control by sending control commands, downloading process data, moni- toring the process, and handling exception events. From the perspective of message transmission, the hierarchy in Figure 10 can be classified into three standard levels, as shown in Figure 11:
1. Production message standard: a standard for obtaining / sending production information from / to low-level computers. The production information can be a routing, a real-time statistics, etc. Usually the information is not directly related to the equipment control.
2. Control message standard: a standard for controlling the equipment logically. The control message includes commands, the parameters associated with the command, and data. SECS- II is a standard that can fulfill such a need (Elnakhal and Rzehak 1993).
3. Virtual machine: a mechanism that converts the logical commands based on the control mes- sage standard, e.g., SECS-II, to commands format that can be accepted by a physical machine. Figure 12 shows that a virtual machine is laid between a host that delivers SECS-II commands and a real device that receives and executes the commands in its format. A virtual machine is therefore responsible for bridging the message format gap.
The development of virtual machines reduces the problems of incompatibility in an automated system, especially when the automated equipment follows different communication standards and protocols. It also enables the high-level controller to focus on his or her control activities, such as sequencing and scheduling, rather than detailed command incompatibility problems. Further infor- mation on virtual machines and automation can be found in Burdea and Coiffet (1999).
Tool Perspective Environment
Modern computer and communication technologies not only improve the quality, accuracy, and time- liness of design and decisions but also provide tools for automating the design and decision making processes. Under such a tool perspective, human experts focus on developing tools, then users can apply the tools to model and resolve their automation problems (see the right-side diagram in Figure 13). In contrast with traditional approaches, difficulties occur only during the tool development. For traditional approaches, ‘‘experts’’ are the bottlenecks in design projects. They must understand not only the problems but also the applications and limitations in practice. In practice, the costs of looking for and working with the experts are the main expense. For tool perspective, the cost of training the users becomes the major expense. However, frequent environment changes usually result in the design of a flexible system to respond to repeated needs to modify and evaluate the existing system. Mod- ification and evaluation needs can be fulfilled by rapidly applying computerized tools, which provide relatively high flexibility in design. Some researchers have noted the importance of modeling man- ufacturing systems with the tool perspective. A sample survey is presented in Table 2. The three modeling concerns ((1) conflicts among designers, (2) constraints in physical environment, (3) the information flows in manufacturing systems) are part of the criteria in the survey table. It is found
that the concerns are not addressed by the first four models. The approach of the first four models still follows the traditional approaches, except for the second model. In the following paragraphs, two automated, integrated tools for manufacturing system design are introduced to present the above concept: facility description language (FDL) and the other is concurrent flexible specification (CFS).
Facility Description Language (FDL)
FDL provides an integrated modeling tool for distributed engineers working on various aspects of manufacturing system design (Witzerman and Nof 1995, 1996; Lara et al. 2000). FDL is implemented in a 3D emulation environment, ROBCAD (Tecnomatix 1989), which provides a dynamic and phys- ically visible (comparing with ‘‘iconically visible’’ in simulation animation) CAD environment. Therefore, all the materials, operations of robots and material handling facilities, and machines are shown by their true physical relationships.
In the framework of FDL, the manufacturing systems are modeled by computer information and graphics. Various information teams are included in the model that are not seen in traditional models (Witzerman and Nof 1995):
1. Organizational relationship among facility components
2. Specification of working locations for robots
3. Flow of parts and materials through the facility
4. Location of personnel and equipment aisles
5. Control relationships between devices
6. Designation of sensors and their targets
These information items are supported by the modeling functions of FDL (Table 3). An example of modeling an aisle by an FDL function is:
The model in FDL then becomes a list of syntax. The list of syntax triggers the ROBCAD program to construct a 3D emulation model (see the model inside the window of Figure 14). Figure 14 is an example of FDL in a ROBCAD system. The upper left window is used to input the information of the system, including the geometric information of facilities, material flow information, and material flow control information. The lower left window is used to show the output information (e.g., a collision occurring to the robot during the material handling). In addition, FDL provides a reconcil- iation function (see the right function menu in Figure 14). Therefore, all the control and physical conflicts on the manufacturing systems can be resolved according to the built in algorithm. The reconciliation function may change the positions of robots or machines to avoid the collision or unreachability of material handling. Recently, FDL / CR has been developed to provide knowledge- based computer support for conflict resolution among distributed designers.
Because FDL provides such direct syntax specifications, the designers can use the syntax to model and develop their subsystems. When the designers are in different locations, their subsystems can submit input to the host ROBCAD system to construct the entire system, then use the reconciliation function to adjust the subsystems if conflicts occur. Therefore, the cooperation of designers in dif- ferent locations for different subsystems can be achieved in FDL. In the FDL working environment, two types of information are exchanged among the designers: (1) the design based on FDL syntax and (2) the operations of the facilities described by a task description language (TDL). TDL represents the control functions of the facilities. Thus, not only the models of the material flow but also the control information are presented and shared among the designers.
Concurrent Flexible Specifications (CFS)
By specification, engineers describe the way that a system should be constructed. Concurrent, flexible specifications for manufacturing systems, in other words, are provided by several engineers to model manufacturing systems with flexibility to design changes. In a manufacturing system, there is a physical flow of raw materials, parts, and subassemblies, together with an information and control flow consisting of status (system state) and control signals. The control and status signals govern the behavior of the physical flows. In order to simultaneously achieve optimal capacity loading with maintained or increased flexibility, an exact definition of material and information flow becomes necessary. This approach is followed by CFS modeling (Furtado and Nof 1995). The specification should represent not only the logical structures of different functions but also their logical connec- tions, such as the structures of material and information flow in a manufacturing cell (Csurgai et al. 1986).
Another important requirement of specification is the ability to define precisely the real-time behavior of the system. In a real-time system, many of the inputs to the system are signals that indicate the occurrence of events. These inputs do not pass data to the system to be processed. Generally they occur in streams over time and their purpose is to trigger some process in the system
repeatedly. Furthermore, many systems are made up of subsystems, any of which may be active or nonactive at a particular time during system operation. For this reason, the treatment of timing is an essential element of the specification.
To integrate the above requirements, tools that can incorporate material flows, information flows, and control signals are required. Therefore, different representations of specifications through different tools should not be independent or mutually exclusive but should support each other by forming a concurrent, comprehensive specification of the system. For manufacturing systems, the specification of functional and real-time logic is important because these attributes describe what and how the system is executing. This information is necessary to determine how the processes are to be imple- mented with physical equipment. By utilizing two complementary representations, both these aspects of system behavior can be specified concurrently.
Data / control flow diagrams (DFD / CFDs), which are enhanced with real-time extensions when used in conjunction with Petri nets, provide a suitable framework for concurrent specification of functional and real-time state logic. The main reason for this is the ability to maintain identical structural decompositions in both representations at all levels of detail in the specification. This model is accomplished by maintaining identical partitioning of processes in both specifications.
With DFD / CFDs, partitioning is accomplished by hierarchical decomposition of bubbles that represent processes or tasks. An identical hierarchical partitioning can be created with Petri nets by representing processes with subnets at higher levels and then showing the detailed, internal net at lower levels of detail. The DFD / CFDs provide a process definition model of the system, while the Petri nets provide a process analysis model for the study of real-time state behavior. Even though object-oriented modeling is becoming a more popular technique of system design, data / control flow diagrams are still an acceptable technique in our case study. Researchers have proved the possibility of transforming data flow diagrams to object models (Alabiso 1988).
Both these techniques are realized by two software packages: Teamwork and P-NUT. Teamwork is a computer aided software engineering (CASE) tool family that automates standard structured methodologies using interactive computer graphics and multiuser workstation power. P-NUT is a set of tools developed by the Distributed Systems Project in the Information and Computer Science
Department of the University of California at Irvine (Razouk 1987) to assist engineers in applying various Petri net-based analysis methods.
Agent-Based Control Systems
Early agents were defined for distributed artificial intelligence, which includes two main areas: dis- tributed problem solving (DPS) and multi-agent systems (MASs). DPS focuses on centrally designed systems solving global problems and applying build-in cooperation strategies. In contrast, MAS deals with heterogeneous agents whose goal is to plan their utility-maximizing coexistence. Examples of DPS are mobile robots exploring uncertain terrain, and task scheduling in manufacturing facilities. Both can be operated with centralized programs, but in relatively more distributed environments they are usually more effective with autonomous programs, or agents. Examples of MAS are collaborative product design and group behavior of several mobile robots.
Recent research has explored the concept of autonomous agents in control. An agent is a com- puting system that can autonomously react and reflex to the impacts from the environment in ac- cordance with its given goal(s). An agent reacts to the environment by executing some preloaded program. Meanwhile, there is an autonomous adjustment mechanism to provide a threshold. When the environmental impacts are higher than the threshold, the agent reflexes; otherwise it is intact. An agent may seek collaboration through communicating with other agents. The communication among agents is regulated by protocols, structure of dialogue, to enhance the effectiveness and efficiency of communication.
An important difference between autonomous agents and other techniques is that an autonomous agent evaluates the rules that it will perform. It may even automatically change its goal to keep itself alive in a harsh environment. Autonomous agents have been applied in many control systems, in- cluding air traffic control, manufacturing process control, and patient monitoring.
Usually an agent functions not alone, but as a member of a group of agents or an agent network. The interaction and communication among agents can be explained by the analogy of organizational communication. An organization is an identifiable social pursuing multiple objectives through the coordinated activities and relations among members and objects. Such a social system is open ended and depends for its effectiveness and survival on other individuals and subsystems in the society of all related organizations and individuals. (It is actually similar for both human societies and agent societies.) Following this analogy, three characteristics of an organization and of an agent network can be observed (Weick 1990):
1. Entities and organization
2. Goals and coordinated activities
3. Adaptability and survivability of the organization
Five motivations have been observed for organizational and agent network communication (Jablin 1990):
1. Generate and obtain information
2. Process and integrate information
3. Share information needed for the coordination of interdependent organizational tasks
4. Disseminate decisions
5. Reinforce a group’s perspective or consensus
These five motivations can serve as a checklist for developing protocols. One of the most influential factors affecting interpersonal or interagent communication patterns among group members is the characteristic of the task on which they are working. As task certainty increases, the group coordinates itself more through formal rules and plans than through individualized communication modes. There- fore, the interacting behaviors and information exchanges among agents have to follow interaction and communication protocols.
Although different agent applications will require different agent design, five general areas have to be addressed:
1. Goal identification and task assignment
2. Distribution of knowledge
3. Organization of the agents
4. Coordination mechanism and protocols
5. Learning and adaptive schemes
Comments
Post a Comment