How autonomous are today’s industrial systems?
Wilhelm Wiese ABB Global Industries and Services Bengaluru, India, firstname.lastname@example.org
Can a production facility’s level of autonomy be compared to the levels of autonomy now generally accepted for vehicles? Far from being an apples-and-oranges comparison, these two application areas share a world of similarities – thus shedding light on the meaning of autonomy itself.
Everybody’s talking about autonomous systems as they apply to cars. But what’s the status of these systems in industry? While the US National Highway Traffic Safety Administration (NHTSA) has established very clear definitions of the five levels of autonomy , similar definitions in the field of industrial automation are yet to be set. Nevertheless, these two application areas can be readily compared.
But just to avoid confusion, it must be clear that automation and autonomous systems are two substantially different animals. Simply put, autonomous systems are characterized by the ability to act independently of direct human control →1, whereas automated systems are not [2,3].
When it comes to autonomous systems in industry, a fair and achievable objective is to target NHTSA’s Level 3. In terms of engineering, when translated into an industrial control system, this can be defined as a system that can itself perform all aspects of a configuration task. Therefore, it identifies the most suitable configuration based on insights it gains from a global data pool of optimized device settings that consider an innumerable combination of connected devices, the industrial domain, and environmental data such as climate conditions.
In terms of operations, Level 3’s industrial counterpart can be described as follows: The human operator must be ready to take back control at any time when the autonomous system requests the human operator to do so.
In both, engineering and operations cases, the objective is the same: to eliminate the need for human intervention through the increasing application of machine learning →2. Technically, this requires significant changes at the control layer of an autonomous system, because the control layer needs to have a holistic plant view.
In view of these circumstances, it can be expected that artificial intelligence will change today’s control paradigm from signal marshalling to process data analytics, from feedback loops to prediction, and from process calibration to self-optimization.
Looking ahead, engineering, operation and control will merge in tomorrow’s autonomous systems into a continuous cycle of self-learning algorithms that will enable process and plant optimizations we can hardly imagine today.