Self Learning Control Systems

The development of autonomous control systems in the present-day world

All of us know about the rise of the self-driving cars, it leaves us completely captivated. If this is possible, what else could be achieved? We’re living in the most affluent era of history with the greatest advancements in technology the world has ever seen. Could we live on Mars? Are flying cars a possibility? Everything is possible. Once we give a man-made system the ability to think, learn and improve, the possibilities are limitless.

Control Systems are systems that are used to maintain a desired result or value. The goal of introducing learning methods to control systems is to broaden the scope of conventional systems and enhance the simplicity of living. The developement of integrated circuits, and the advancement of microprocessors and computers, together, enabled control systems to be developed. Since they were cheap and were able to control a wide range of processes, they spread across to mundane processes such as the domestic washing machine. Self learning behaviour is a very desirable characteristic of advanced control systems so they continue to perform well under dynamic conditions of the environment, without external intervention. This compels the system to tailor itself to the changes caused in the surroundings. To learn from past experience is a covetable feature, and, to make decisions based on these experiences is the primary focus of the system.

These systems can engineer their output based on their stimuli more successfully since it has the ability to recogonise how much it can improve. The type of action taken by the machine is dependent upon the nature of the system and the type of learning system implemented. For example, a robot may need to change its visual representation of the surroundings after learning of new obstacles in the environment.

The way toward learning might be accomplished disconnected, utilising incomplete data illustrations of the watched or expected processes. The data got from the assessment of the control systems conduct might be numeric, or heuristic. In this way, self learning control system ought to have the capacity to deal with representative date and numeric information. Where can learning be utilised as a part of the control of systems?

Figure 1 Block Diagram showing a feedback control system

(Modified from figure 1.5 in Reference in 1 and figure 1.4 in Reference 2 )

Figure 1 illustrates a basic central heating control system. It requires a manual input for the intended temperature of the house. However, if a precise self-learning algorithm was added, the manual input would be redundant. It would automatically be able to recogonise the temperature required by gauging the temperature outside, time of the day, number of people in the house etc. This would avoid so many tiny annoyances such as freezing while entering the house, or disputes over what temperature it should be set at. We’ve all been there, haven’t we? The cherry-on-top would be the fact that the system would get better every second. Every time, it would learn the preferences of the members living in the house. It would learn at what temperature they perform, sleep, eat and play. It would recieve data, process it and improve based on the feedback. It would achieve perfection.

As said, learning assumes a fundamental part in the self-governing control of systems. There are numerous regions in control where learning can be utilised to our advantage, and these requirements can be described as:

1. How to infer new plant models.

2. Finding out about the earth.

3. Figuring out how to change certain controller parameters to upgrade execution.

4. Adapting to new plan objectives.

In many control issues, our insight about the issue is deficient. Either, there is some procedural mistake, or the objectives are not well defined. A sub optimal system would be obtained. This would lead to the system making frequent errors. The industrial control systems in the present-day world are at a higher danger of obstruction by hackers. This is a result of their connection to the web. While digitisation helps in automating these systems, it additionally gives an access for hackers to access and meddle with them.

If the self-learning system has an excellent design, a well-thought-of algorithm, and a solid infrastructure, with regular maintainance it could eventually lead to an impeccable and foolproof process making the lives of the people involved effortless. However, to be able to achieve perfection, we have lots to learn, develop and test. On the other hand, the industrial sector is rapidly implementing self-learning control systems as they aim to automate monotonous tasks. However, the insurance of these systems is essential because if they fall due to hacking or faulty algorithms, the dependent systems will become futile and the collective consequences can be gigantic.

References

1. Bolton, W. 2002. Control Systems. Oxford: Elsevier Ltd.

2. Sinha, N. 1998. Control Systems. 3rd ed. New Delhi: New age international Ltd.

3. Nguyen, D. Widrow, B. Neural Networks for Self-Learning Control Systems. IEEE Control Systems Magazine : IEEE. 1990. p. 18-23.

4. Guez, A. Eilbert, J.L. Kam, M. Neural network architecture for control. IEEE Control Systems Magazine: IEEE. 1988. p. 22-25.

5. Antsaklis, P. J. Intelligent Learning Control. IEEE Control Systems Magazine: IEEE. 1995. p. 5-7.

6. Albertos, P. Picó, J. Sala, A. 1997. Learning control systems: An overview. Valencia: Department of Systems Engineering and Automation.

7. Cyware. 7 November 2016. 3 Cyber Attacks that rocked Industrial Control Systems. Available from: https://medium.com/@Cyware.com/3-cyber-attacks-that-rocked-industrial-control-systems-c81e98001750 .

This work is licensed under a Creative Commons Attribution 4.0 International License.