Automation & Robotics          




Automation may be defined as the set of methods to achieve automatic control of a manufacturing process through successive stages without direct human intervention 22. These methods may incorporate aspects of electronic, mechanical and computer-based technology in order to operate and control the overall system. It can be classified into three main classes:

a) Fixed automation, a relatively inflexible system used for high production volumes where the sequence of operations is given by the equipment configuration;

b) Programmable automation, where the sequence of operations can be changed to adapt to product configuration in smaller batch production jobs thus allowing flexibility with respect to product changes;

c) Flexible automation that is an extension of the previous category to incorporate much greater changeover flexibility and continuous adjustment to a variable product mix.



Virtual Manufacturing and Robotic Cell Design (30)

To meet the pressures of time-to-market, factory layout and process design software must be as fast and flexible as 3D CAD product design packages. Virtual manufacturing, the digital factory, or factory simulation software is a response to this challenge. Current packages offer a range of tools to handle automatic assembly sequences, for instance by capturing data on:

¤ Human motion/accessibility ranges during assembly tasks,
¤ Potential collisions for given equipment placement and operations, and
¤ Optimum work flows for parts of different sizes.
For simple analysis without spatial constraints, 2D and 3D icon-based products are probably sufficient. More sophisticated software commercially available today allow the use of non-standard machines with many different attributes. Work cell virtual fly-through and assembly task ergonomics can be incorporated into the virtual manufacturing picture as well as detailed robotic control programming or CNC simulation software.
Existing technology allows the simulation of almost any aspect of product and process design at varying levels of complexity and success. For instance, human-machine interactions can be represented through ergonomic models for training purposes or to evaluate mechanical and global performances. The former evaluation may make use of structural simulation, collision detection, robot controllers or CNC program simulation. These simulations may use Monte Carlo techniques, continuous or discrete simulation approaches.

Robot cell design is a particularly difficult task that requires 3D representations because 2D views may be misinterpreted. The are several requirements to meet:

¤ Predict robots or automates movement,
¤ Avoid interference between moving equipment or automated devices
¤ Determine the optimum placement for maximum reach
¤ Design within confined spaces with high risk of collision
¤ Identify possible production bottlenecks




¤ Minimise the length of total production cycle.
¤ Consider safety requirements (including equipment), rules and regulations
¤ Avoid injury risks
¤ Consider training needs for operators with no risk for either trainer or trainee.

Faster work cell layout design (and re-design) tools can minimise the required engineering design effort and reduce down time associated with on-line programming. Costly mistakes can be avoided, for instance by resetting equipment's reach and collision detection parameters to meet scheduling requirements. More importantly, the optimisation of the work cell can be evaluated and programmed off-line, allowing the system to continue operating during the planning phase. The resulting requirements for the supporting 3D simulating system become obvious (31):

(a) It must be a powerful and user-friendly tool

(b) Accurate models of real industrial must be easily accessible as well as mechanical modelling tools (for example, through a library)

(c) Automated tools placement must be represented to test reach, tool orientation and highlight any changes in cell configuration

(d) Other automated plant components such as conveyor belts, part feeders, CNC machines may also need to be included in the simulation model

(e) A calibration facility must be available to increase model accuracy (32)

(e) A collision detection feature must be included in the software; integration between production scheduling routines and the simulation model is highly desirable

(g) Off-line programming facility for subsequent downloading is a cost-effective feature

(h) Simulation, calibration and off-line programming of industrial automated equipment should be possible, ideally using a standard low-cost desk or laptop PC.





It has been argued that the pressure on new developments in the field of command and control is largely due to the shortcomings in the automation field (33). In today's multi-product manufacturing environment, human and material resources must be managed in a rational and concurrent manner.

The optimisation of flows throughout the production system is a main priority as well a dynamic allocation of available material resources to minimise costs and delays in production and distribution.

The flexibility requirements impose a dynamic adaptive system that responds well to internal and external shocks such as equipment maintenance requirements and market demand fluctuations, respectively. Industrial needs require a command and control system that can be easily responsive to its fluctuating needs in real time while remaining capable of evolving and being reconfigured with relative ease. Modular command and control systems seem particularly adapted to the reactive and dynamic needs of industry.

The problem facing command and control system designers is to develop a multi-agent architecture with a communication network, which links the various system agents. These agents may be physical or abstract with relative levels of autonomy and capacity to affect their environment as well as change their own behaviour. For that, they dispose of a partial representation of their environment obtained through embedded means of perception and communication.

The resulting behaviour rules observed are a consequence of the agent's own observations, level of knowledge and interactions with other agents. If the agent's knowledge is to be shared with other agents in real or virtual form and the amount of redundancy are important design considerations34. In short, the agents have a social tendency vis-à-vis the entire system as well as an individual tendency centred around its own operating rules. All agents have characteristics such as:



a) Intentions and objectives as the agent moves towards a pre-defined set of goals given the means available

b) Rationality as an agent will select the best course of action based upon a set of evaluation criteria

c) planning ability in that an agent will co-ordinate its actions with other agents and will plan how to attain its own objectives

d) Adaptability with respect to the environment and the other agents with which one agent interacts; e) intelligence if the agent has all the characteristics listed in a) to d)

Complex intelligent agents have an explicit representation of their environment, may keep track of past performance and events, and are probably small in number. Most systems also contain a large number of reactive agents with limited protocol and communication skills and no explicit representation of the outside world. Not all agents are required to be individually intelligent for a command and control system to be considered intelligent. The three objectives of a multi-agent system can be summarised in three words: communication, control and organisation 35. There are various design options in terms of communication. Agents may be unable to communicate directly with each other but go through a common blackboard. They can exchange asynchronous messages, be spatially distributed with a large degree of autonomy, and rely on a set of pre-defined production rules placed on an inference engine or follow recent trends towards the more abstract holonic or neural communication system.

Command and control architectures can be classified depending upon the forms of communication and control used throughout the system. These may be a meta-object agent, supervisory agents, cell agents, product agents and finally resource agents, the latter being non-intelligent in classic workstations 36. The control architecture may be centralised (with and without pre-planned production schedules), hierarchical, co-ordinated, distributed and finally distributed and supervised. In the case of the co-ordinated and distributed architectures, intra-level communication links exist with or without a global supervisor (37).



The choice of the most appropriate command and control architecture depends on several process design parameters. These include the total number of machines, required response times, the configuration of the production system itself (e.g. use of cells, number of identical and non-identical cells), communication requirements among machines and/or cells, trade-off between intelligence and reactivity requirements, and the existence of a production schedule that can be used for pre-planning purposes.

The ideal architecture is reactive, robust, fast and reliable (33). Once it is provided with product type, quantity, delivery date, optimum starting date and product priority code, the system must be able to locate the necessary resources (inputs, machines, etc.) and evaluate their availability in time and space.

In summary, the command and control function is in a central position linking Production Planning and Scheduling and Automation and Robotics. It transmits scheduling orders and their due dates to the operating equipment in the production plant after identification of the resources required for particular tasks. This function requires translation and interpretation between two different languages.

What makes this function particularly critical is the fact that it can evaluate equipment availability. In the case of malfunctions for instance, it can adopt remedial measures in real time, one of which may be to demand an alternative task/resource scheduling routine that will temporarily eliminate the faulty machine from the circuit.

The speed of the response time is an important measure of effectiveness of the command and control function. Alternatively, the quality of the command and control function response may be considered an important alternative to speed. In this case, once faced with a resource availability problem, the system makes an extended analysis (without time consideration) and capitalises on that new knowledge. In that situation, the command and control system is often based on an expert system and the rules base consists of acquired working knowledge.




Once the best solution for a particular problem is obtained, it will be reused in a similar situation. Evidently, these two philosophies cannot be applied simultaneously to the same type of systems. The second type in particular is reserved for systems when response time is irrelevant (as in the case of long production cycles) but it is imperative to apply the best remedial solution for a particular problem because of costs, for instance. As discussed in detail above, there are three broad choices to automate the command and control function:

¤ A unique application manages all, an ideal choice for small systems because of the risk of an explosive increase in operating time for large systems.

¤ A multi-agent application in which each agent is assigned to a particular task, a much more flexible alternative. There are many architectures of this type, the choice of which depends on the particular configuration of the operational system.

¤ A multi-expert system that is a particular application of the multi-agent system where each agent is an expert system. This particular choice is almost exclusively applied to systems adopting the second philosophy where the quality of the response is more important than response time.

In conclusion, the command and control module plays a central role in the CIM chain. It simplifies the shop-floor manager job by automating some of the managerial functions. It provides a reliable link between scheduling and numerically controlled automated machines by translating and interpreting scheduling requirements. Finally, it relies on a high performance computer system for quick reactivity to endogenous and exogenous shocks to the productive system.






Production planning 41 is typically a medium-term activity. Based on demand for the various products to be manufactured, production is allocated to time periods within the planning horizon, taking into account capacity constraints.

The plan will depend on the relative values of set-up costs and of carrying inventory from one period to another. With large set-up costs, longer production runs are preferred. On the other hand, when set-up costs are small, then short production runs are chosen in order to avoid the expense of holding larger inventories. Various complications arise in production planning models.

Since the plans are usually designed for several months in the future, not all customer orders will have been received. Thus, forecasting techniques play an important role in estimating demand. Also, complex products may require work in different departments or on different machines. In such cases, any precedence rules between different operations must be respected. Planning decisions are also affected by the availability of resources such as machine capacity as well as constraints on labour or materials.

Production planning problems operate at a fairly high level of aggregation, and do not account for the movement of work between machines on an hourly or daily basis, for instance. This level of analysis is performed by production scheduling specialists.



Production scheduling is a short-term activity and provides a detailed specification of the work that the machines are to perform, usually over the period of a few days or weeks. In a scheduling problem, jobs must be processed within the time period fixed at the production planning stage, as well as their arrival times and due dates, whereas the order in which the jobs are to be processed has to be determined.

The most basic model requires a single machine to be scheduled. If there are several machines performing the same function, a parallel machine-scheduling problem results. More complex models require several operations at different stages to be performed on the jobs. In a flow shop, every job passes through the machines in the same order. However, in the more general job shop model, different jobs have different machine routings and a job may revisit a machine (in a re-entrant system).

The objective is sometimes to minimise the maximum completion time or the sum of weighted completion times of the jobs. However, in the presence of due dates, the objective may be to minimise other performance measures such as the maximum lateness, the total weighted tardiness or the weighted number of late jobs.

With the increased use of flexible and automated machinery, scheduling problems differ from the classical ones mentioned above. More precisely, the assumption in classical scheduling is that each type of operation is performed on a dedicated machine or group of machines, whereas in modern production systems, there is often a choice of which machine or tool to use for an operation. Thus, scheduling problems require an allocation of operations to machines as well as the determination of the operations processing order for each machine.

The investment in the provision of an automated production system is often substantial. To reap the benefits of such heavy investment, a high level of utilisation of the machinery should be achieved. For the repetitive manufacture of high volume products, therefore, a common objective is to maximise the throughput of the products.






An important feature in the design of a manufacturing system is the positioning of machines on the shop floor. The layout of machines will be constrained by the available space, and the need to access the machines, either by operators, by a transportation device to move work between machines, or for maintenance to be carried out. Ideally, a layout should position machines close together if there is a high flow of work between them.

Many layout models consider the problem of assigning machines to predetermined locations on the shop floor. If the objective is to minimise the flow of work between a machine pair multiplied by the distance between these machines in their chosen locations, summed over all machine pairs, then the resulting model is a quadratic assignment problem.

Other models lead to graph-theoretic formulations. By assigning a weight to each machine pair to indicate the desirability of locating these machines in adjacent positions, the problem of maximising the total weight of adjacent machines becomes a planar sub-graph problem.

More specific layout problems arise when the layout must conform to particular type. Types of layout include single row, multiple row and loop, where the single row can take the form of a linear, U-shape or semicircular layout. For a linear layout, it is often desirable to sequence the machines so that backtracking to previous machines is minimised.

However, for a loop layout in which the machines are arranged around a conveyor, it is appropriate to sequence the machines to reduce the number of loops required to manufacture the products.

A particularly important decision concerning plant layout is the use of manufacturing cells. In this case, the manufacturing of a particular part type requires a sequence of operations to be performed by any machine of a given type. The cell formation problem consists in grouping machines into cells and assigning operations to machines as to minimise inter-cell traffic, for instance (43).



The application of concurrent engineering principles during the first drawings meant that the modules would be given general information about the case study and that the staff would formulate their own set of preferences to select the best design for that FMS.

Given the wide variety of modelling approaches used in process design and the interdisciplinary nature of most planning groups and in this project, the choice of a common and easily understood modelling tool is a major concern. Graphs, 3D Design & Simulation are selected as modelling tools for the FMS design analysis for some reasons:

¤ It is commonly used in disciplines dealing with flow optimisation such as logistics, production planning and plant lay-out and a modern standard in automation and robotics, or command and control;

¤ It allows the joint evaluation of module solutions;

¤ It provides a smooth convergence to a global optimum solution, if one exists.

For one case study, various design features and all feasible combinations have to be assigned, the set of logically consistent options across modules define the best feasible solution, if one exists. Each of the modules is based on experience or modelling results but should take into account preferences from other modules, i.e. optimised solutions.

Internal module preferences may be fully represented in the 3D simulations techniques (such as the A&R - Automation and Robotics module) but be reflected in the total weight placed on a particular design. In some other cases, the complete set of individual module objectives cannot be evaluated in the global context (such as balanced workload, long production runs and location of WIP - Working in Process storage for the P&L - Production & Layout module). In this case, modules may simply assign a global weight to the set of design features available before providing its own contribution (this is the case for the P&L and the M&L - Maintenance & Logistics modules shown bellow in the FMS designs). In some other cases, modules may not assign any intermediate options.

Although this approach provides several feasible solutions, it has the disadvantage that the amount of information increases from top to bottom. Subsequent runs of the same problem in a different order may alter the order of preferences and the ranking of the solutions. The selection of the most desirable module consultation order is one of the simulator's decision variables in its search for optimality.

For example, the C&C Command & Control module may be consulted at the bottom end rather than at the top of the tree because it requires a significant amount of information from other modules. This includes the number and type of machines proposed, the use of cells for production (or not), scheduling, maintenance and logistics requirement.

The M&L module requirements will have an impact on plant layout and equipment choice. And so on. In other words, there is an important role to be played by the 3D simulator. Otherwise it must be decided the order in which it visits the modules, how to obtain global feasibility, how to impose additional constraints on individual modules if necessary and finally how to move towards global optimality.

In a case study, many feasible solutions are identified, a number of which being clearly superior to the others thus providing an important first step to discuss optimisation as well as the best approach for the general case. The drawings below are a summary of the strategic decision variables selected by each of the four modules and the set of solutions assigned to the various options. The complete set of options and options for the case study following the analysis and approval were simulated to propose a Generic FMS design.











The generic model is applicable to a vast chemical and cosmetics products lines using sterilisation by irradiation and ADSR – Automatic Distribution Storage Retrieval Systems.

Seções Principais Domínios Pesquisa Industrial Tecnologias Projetos Materiais Informações