We assume that scheduling is the activity of creating a schedule, that is a detailed, short-term plan stating a sequence of tasks on each resource together with a start- and end time for each task. A schedule is created with respect to one or more optimization criteria. A schedule may be created in advance (off-line scheduling), created or revised in real-time based on on-line monitoring information (by reactive scheduling). It may be determined by centralized software or by software systems which are designed in a distributed manner (distributed scheduling).

Further we assume that the scheduling function does not encapsulate the derivation of basic control decisions for the production, transportation and storage units (e.g., directing a specific work-piece at a cross-road or picking a pallet from a buffer and putting it to the load place off a manufacturing cell) from the schedule. That is, while scheduling is concerned with the generation of a detailed plan, the execution of the plan is left to the shop floor control system or (in manual environments) to the workers.

If scheduling is interpreted in this way, one main difference between MASCADA and scheduling is the scope: MASCADA will develop a manufacturing control system which covers both the short term plan generation and the plan execution aspects. (Clearly, If scheduling is interpreted in a broader sense, i.e. covering these both aspects, then there is no difference in scope between scheduling and MASCADA.)

Please note that considerable different interpretations of the term ‘execution’ exist in different communities. For example, the Manufacturing Execution Systems Association (MESA International, http://www.mesa.org) uses the term execution to describe activities at the short-term planning, scheduling or dispatching level. They interpret these activities as the execution of the more general, mid-term plans generated by MRP planning. The level below scheduling is there defined as control (in the sense of NC / CNC machine control). Control of transport systems is not considered there.

MASCADA uses the term execution to describe activities at the control level, especially for the control of the transport and storage system.

A difference between traditionally scheduling approaches and MASCADA lies in the granularity (or degree of detail) of the generated short-term plans. Schedules in the classical sense are very detailed, i.e. they determine sequences of single workpieces or batches on a resource as well as time points or windows for the start and the duration of the operations on them. The granularity of short-term plans in MASCADA may vary, depending on the properties of the considered production system, from detailed plans (like classical schedules) up to broader strategies or partial plans for specific resources. For example, in some industrial applications the frequency of disturbances is such that specific schedules are not ever set but rather, broad strategies are provided as guidelines to more local decision making. This would in fact be the case at the Daimler-Benz Sindelfingen Paint Shop.

A further difference lies in the manner in which the generation and execution activities interact:
A) Steady Operations: Traditionally, schedule execution in the presence of a schedule (as defined above as a detailed plan) is seen as the task to follow the schedule by all means, e.g. as in Computer Integrated Manufacturing (CIM). This is not the case in the MASCADA approach: Mascada control systems will know about the goals behind the scheduling, and use the available schedules as useful information to control the production system. In situations without disturbances, this probably means that the control system follows the schedule and makes the decisions which the scheduler omits.
B) Disturbance Conditions: When disturbances render the schedule inapplicable, the control system will aim to achieve the system goals as good as possible. Under these circumstances, the control system takes its decisions independently until the scheduling system provides an up-to-date schedule and may not, for example, attempt to follow the schedule at all. The Mascada control system does not need to understand the internal workings of the scheduling system, but will know about the relevant and available schedules (i.e. outputs of the scheduling system), and knows about the production goals which govern the operations of the scheduling system.

In the case of industrial applications where only broad strategies are provided as guidelines to the local decision making, this decision making would be done by the Mascada control system at the control level, i.e. by agents which are directly connected to (perhaps, physically located at) the PLC’s.

To capture the fact that Mascada control system may work in conjunction with both preset schedules and broad strategies, we use the term short term planning (instead of scheduling) to cover the Mascada control system actions for both of these possibilities.

We summarize the scope of MASCADA and its relation to scheduling and other Production Planning and Control tasks in the diagram below.

Disturbance handling in MASCADA consists mainly of two tasks which are closely related to the two main levels of the MASCADA control system. At the plan execution level the system takes its local decisions having in mind the objectives of the production system but possibly contradicting the existing, no longer valid plan. At the planning level, a new or revised plan is generated. When this re-planning is performed depends on the nature of the production system and the granularity of the plan. For example, to avoid nervousness of the production system a re-planning may be delayed until more information about the disturbance is available. For very short disturbances it may be not necessary to change the plan, the system can rely on the abilities of the control level to resolve the problem.

The explicit nature of the disturbance examination at the planning level is based on an understanding and classification of the known classes of disturbances and the manner in which they typically impact on operations. By understanding this impact better, it will be possible to influence short term planning such that the manner in which operations are planned is such that performance, in terms of production goals, is most robust to possible disturbances. The method for influencing these plans will be via a weighting of agent decision making mechanisms. It should be emphasized that such an approach is complementary (rather than an alternative) to an effective reactive approach to managing operational disturbances. It is envisaged that the type of decision variables relevant to agent decision making in the context of this proactive form of disturbance handling will be, for example:

• point of routing commitment
• locations & size of buffers
• allocation of slack to "vulnerable" resources.

A rudimentary example of this ‘proactive’ approach to disturbance management can be found in (Duffie & Prabhu, Journal of Manufacturing Systems, V13, 1994).

The existence and incorporation of decision rules is clarified with respect to the two system levels of the MASCADA control architecture. First, decision rules at the control level are examined, after that comments to ‘decision rules’ at the planning level will be given.

Decision rules at the execution level
Typically, decision rules are used at the plan execution level (control level) at manually controlled shop floors. The activity of controlling a floor by such rules is often referred to dispatching. In some cases dispatching replaces scheduling and schedule execution completely, then it is often combined with load-oriented order release methods, which are the lower end of the specific MRP planning hierarchy. In other cases it is used as a work around or recovery method in schedule execution if the given schedule became invalid due to disturbances.

For example, such decision rules are currently used by the operator personnel in the main test bed application. They are executed by inputs to the supervisory control and data acquisition (SCADA) system which directly controls the PLC’s. For typical rules, macros are implemented within the SCADA system.
It is explicitly an aim of MASCADA to make operator decision rules at this level obsolete. The introduction of local, intelligent controllers at the plan execution level implies that explicit control of production and transportation resources by the operators will no longer be necessary.
Nevertheless, the current decision rules at the control level have to be analyzed for the following reasons:

1. For the detailed simulation of the current production system, which has to be performed according to the project plan to get a basis for comparison of the system as is currently controlled with the system controlled by the MASCADA control system, it is necessary to map current decision making behavior of the operators. We will not assume that we can collect each single rule operators have in mind, but we will analyze typical rules as e.g. implemented as macros.
2. To build intelligent local controllers it may be useful to provide them with knowledge on the current problem solving (e.g., disturbance handling) rules. This may influence the design of the agent system, i.e. cause specific negotiation protocols or help define which information should be regarded in local decision making. Nevertheless, it is not intended to replicate operational decision making capabilities.

Decision rules at the planning level In some production systems the frequency of disturbances is such that specific schedules are not set but rather broad strategies are provided by the planning level as guidelines to more local decision making. The main test bed application, the Sindelfingen painting center, is such a system. The strategies used there do not concern single work-pieces but rather about ‘streams’. They are not stated by the operator personnel, but by the operational management. Strategies are on more abstract level, and only partly based on technological constraints, i.e., consider also workforce distribution, political reasons, labor unions influences etc.
Examples for strategies are given at the WP1 deep analysis of the Sindelfingen case:

• "70 percent of the orders should be sent to units DL1- DL3 and 30 percent to DL4.",
• "If utilization of the first repair preparation unit is too high, than ten percent of the orders have to be shifted to the second repair preparation unit.",
• "Cars with colors which are known for their high efficiency are to be concentrated at the beginning and the end of each shift. (This is to motivate the workers at the beginning and to compensate their fatigue at the end.)".

Such strategies may be seen as decision rules at the planning level, it should be noted that they are considerably different from decision rules at the floor level. Hence we do not use the term decision rule but the term strategy which we see as a sub-concept of plan in our terminology.
A Mascada control system has to accept such strategies as an input and derive local policies for the sub-units, to be implemented by local control decisions, causing the ‘streams’ in the global strategies to emerge.
In this sense, the MASCADA control system will not directly implement the strategies, but it processes them in the sense of imposed hard constraints at the planning level.
It is an aim of MASCADA to enable the control system to negotiate strategies with users, to present proposals, and to perform what-if analysis for user inputs. It is one explicitly formulated expectation of the production managers in the Sindelfingen case that a MASCADA system will help them to improve their strategies by predicting system behavior for given strategies and proposing new ones. On the other hand, we have to be aware that the user in specific situations will reject system proposals and state his requirements definitively. In such a case, a plan may be identically to a given strategy.

The scope of MASCADA is the control of flexible flow shops. Flow shops are the most important kind of production organization for industries which produces mass or mass customized products.
Modern automated transportation systems in flexible flow-shops make it possible for each work-piece or batch to reach each production resource at the floor. In this sense, highly flexible flow-shops will get one property of current job shops. There, it is assumed that each job can reach each machine.
To compare both approaches, one has to impose the key property of flow-shops, the high load, to the job shop scenario. Under conditions of heavy load the challenge in job-shop is a scheduling problem, the challenge in flow-shop is a control problem due to the coupling constraints of transportation elements.
Imposing this coupling constraints to the job-shop problem would transform it to a flexible flow-shop problem.