Have you ever thought about the safety and reliability of your flight, the steady supply of electricity generated by the power turbines inside a power station or the required temperature adjusted by the air-conditioner in your home or office?
They all have control systems to achieve the desired purpose. Some of them are more complex than others. For example, the control systems of the Hubble space telescope may have very sophisticated positioning control systems, while the water level float regulator, which is a classical control system has a much simpler control mechanism. The regulator is shown in Figure 1.1. The float detects the water level and controls the value that covers the water inlet in the boiler.Figure 1.1 Water-level float regulator
Nowadays, most systems can be modelled by the mathematical and analytical methods developed over the last two decades. Therefore control engineers can have good understanding of any systems and the desired behaviour can be achieved. Also during the last two decades, the exponential drop in price of computing power made computers widely available. Computers are used in most control systems today.
A computer which acts as a logic decision unit processes input provides output in digital form, ie 0 and 1. It is also known as a discrete-time system. It is easy to see that a system which processes only continuous-time data is called a continuous-time system, that is, it can be represented by mathematical functions.
Moreover, it is common to have a mixture of both logic and continuous systems. A simplified model of a large boiler-generator system which is a typical example is shown in Figure 2.2. While all the generators produce continuous-time data, the computer receives those data and processes them in discrete form. This kind of systems is known as hybrid control system.
Hybrid systems are basically networks of interacting digital and analog devices. The new generation of military air fighter, for instance Euro-fighter, is purposefully built unstable so that the aircraft is more manouverable. This is one of the areas in which the control theory of hybrid systems can be applied. Another area using hybrid systems heavily is computer-aided manufacturing. But due to the rapid development of processor and circuit technology, modern cars, for instance, and even consumer electronics use software to control the physical processes. Hybrid system control has already become an important part of study within control engineering.
A typical hybrid system is represented in the block diagram(figure 3.1).
Almost any hybrid control system can be modelled by this two-layer structured block diagram. The layers communicate through the interface. We take a programmable microwave oven which has different preset programs for cooking food to illustrate the above diagram. The continuous control layer is the power control part and turnable part of the oven, whereas the logic part is the preset program or logic. The analog to digital converter and digital to analog converter are the interface.
As we can see, Figure 3.1 is a highly ordered and hierarchical structure, the logical layer(top layer) issues instructions which are converted into continuous input for the continuous layer(bottom layer). The continuous layer feeds reference values back to the interface which converts them into discrete form. If the output of a system feeds back to its input, it is called a closed loop system, otherwise it is called an open loop system. Both types of systems are shown in figure 3.2.
|open loop(without feedback)||closed loop(with feedback)|
Apart from the continuous control system layer, the logic layer is also an important part of a hybrid system. It usually consists of "IF-THEN-ELSE" statements which can be represented by a finite state machine(FSM), see figure 2.3.
As you may have noticed, a hybrid system provides more flexibility than purely continuous systems, since the logical unit has decision-making ability and planning capacity, because decision-making and planning are basically discrete processes.
Even though hybrid systems are so common today, there is still no reliable theory to integrate logic and continuous systems, despite our having well- developed theories on discrete and continuous systems. For most such systems, the logic part and continuous part are designed independently, and then combined by an interface which is designed for solving the specific problem, or the whole system is analysed as either purely discrete or continuous entities.
For small-scale systems, the logic component can be quite complex and the design task can easily become unwieldy if care is not taken during its design. Due to this reason most of the current CACE works well in small-scale hybrid system modelling but not in large-scale.
There are many people working on this problem, for example, Researchers, such as W.Kohn, A.Nerode and others came up with a new approach to CACE. Their approach is briefly explained in the following. The CACE architecture is based on multiple views. These are shown in figure 4.1. This relationship between user, design engineer and system engineer is needed to avoid the delay of development of a system which is caused by the consistent disagreement between them than by specifying each player's responsibilities.
The level of abstraction modelling is supported by the scenario-based requirement analysis. It is because scenario-based modelling is an easily understood and readily modified way of discussing complex processes. The multiple-agent hybrid control architecture(MAHCA) consists of a variable network of control agents MAHCA is the main architecture to support the integration of the above multiple views, see figure 4.2 and 4.3.
Each agent is formally structured; that is, its behaviour is characterised by a model encoded in logic clauses(IF-THEN-ELSE) in hierarchy. Logic failures occur when the model and behaviour of the agent do not agree with. They are triggered by events affecting the processes under control. This model enables us to detect of logic failures and hence we can develop structural adaptation processes to rectify them. Structural adaptation is accomplished by modifying the logic clauses according to a set of rules, or by creating or deleting agents in the network.
It is difficult to solve multiple-agent control problems for several reasons. In non-cooperative problems, agents may disagree on what is the best solutions. Even in cooperative problems, an optimal solution may be computationally unsolvable. MAHCA overcomes these problems by letting each agent have view of the network so that the combination of these models is a near optimal model.
The Maruti operating system will not be discussed in this article. This was developed by the University of Maryland but if you are interested, there is a web page on Maruti. This methodology has been implemented a two-agent based system which is discussed by Ahmak Kamil's article.
The general idea of hybrid control system has been introduced. As we have seen the modelling and implementation are not as easy as they are thought to be. A lot of study has to be done in order to gain more understanding of hybrid systems, as the application of these systems are enormous.