=Paper=
{{Paper
|id=Vol-2900/WS1Paper1
|storemode=property
|title=An Integrators Perspective on AI-Enhanced Cyber-Physical Systems to Support Flexible Configurable Manufacturing
|pdfUrl=https://ceur-ws.org/Vol-2900/WS1Paper1.pdf
|volume=Vol-2900
|authors=Gash Bhullar
|dblpUrl=https://dblp.org/rec/conf/iesa/Bhullar20
}}
==An Integrators Perspective on AI-Enhanced Cyber-Physical Systems to Support Flexible Configurable Manufacturing==
https://ceur-ws.org/Vol-2900/WS1Paper1.pdf
An Integrators perspective on AI-Enhanced Cyber-Physical
Systems to support Flexible Configurable Manufacturing
Gash Bhullara
a
Control 2K Limited, Waterton Technology Centre, Waterton Industrial Estate, Bridgend, CF31 3WT, UK
Abstract
“AI Enhanced Intelligence” in the context of control systems is a term associated with injecting
Artificial Intelligence into shopfloor systems that incorporate PLC, Robotics or CNC control
systems. Flexible Configurable Manufacturing systems require real-time decision-making
capabilities on the shop floor to be effective. As Human-Robot collaboration is increasing
becoming the norm, it is important that certain decisions are left to those who have in-depth
process knowledge and can be supported by providing relevant information to make process
changes. The data collected from the machines is already being used to optimise the process,
but unpredicted events need immediate responses and standard programming architecture is
not designed for flexible configuration and needs to be modified to embrace new knowledge
streams from AI to respond to every eventuality.
Keywords 1
Collective Intelligence, Enhanced Intelligence, interoperability, manufacturing, Artificial
Intelligence, AI, I4.0
1. Introduction
“AI Enhanced Intelligence” in the context of control systems is a term associated with injecting
Artificial Intelligence into shopfloor systems that incorporate Programmable Logic controllers (PLCs),
Robotics or Computer Numeric Control (CNC) systems. Flexible Configurable Manufacturing systems
require real-time decision-making capabilities on the shop floor to be effective but currently, much of
the processing of the data is carried out on remote servers and therefore introduce time delays which
means that it is not possible to reconfigure production systems on the fly. As Human-Robot
collaboration is increasing becoming the norm, it is important that certain decisions are left to those
who have in-depth process knowledge and can be supported by providing relevant information to make
process changes. Of course, this assumes that operators have had time to become skilled and can make
such critical decisions, so a belt and braces approach is always preferable in case the wrong decisions
are made. The data collected from the machines is already being used to optimise the process, but
unpredicted events need immediate responses and standard programming architecture is not designed
for flexible configuration and needs to be modified to embrace new knowledge streams from AI to
respond to every eventuality.
2. Input signals for AI Enhanced CPS Systems
The development of technologies can often be traced back to the early pioneering developments of
IT or Control Systems. Automating tasks has been the cornerstone of the manufacturing sector to reduce
Proceedings of the Workshops of I-ESA 2020, 17-11-2020, Tarbes, France
©️ 2020 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
labour costs. The replacement of people with machines is nothing new; it has always been about
repeatability, reliability, accuracy and speeding up tasks. Vision systems for example, have been around
for a long time and AI has been used behind the scenes to provide the data required that mimics the
observing operator but with much better accuracy and longer-term performance.
Figure 1: Knowledge Chain of Pulled Data from Shopfloor
A multitude of heterogeneous systems are typically deployed on the shopfloor all supporting the
drive for greater productivity.
If all this data and technology could be filtered into one deployable solution it would clearly
significantly increase production, improve efficiency, and reduce waste for all manufacturing
companies. Figure 1 shows the data paths of incoming knowledge that could be filtered into an
“Enhanced CPS System” (ECPS) and provide vital intelligence for decision making. The added feature
of this knowledge is not just about the process and associated hardware, but additional data pertaining
to the “surroundings” of the production line could also be collated.
This new knowledge is the start point to “Collective Intelligence” [1] that can be “injected” into
control systems to eliminate errors typically caused by imperfections and anomalies in the production
process. The more “connected” an enterprise becomes, the better the chances of achieving the reduction
of costs and increasing profits.
3. Challenges of incorporating AI into Control systems
In Julian Birkinshaw’s paper “The future of firms in an AI world” [2] the increasing role of AI in all
aspects of human interaction within social and business environments is going to radically change.
Every process, every model and every system following on from the Industry 4.0 revolution, will need
a complete overhaul. Systems thinking will need to make way for AI enhanced systems that will cross
many barriers and sectors. So, it is clear that big changes are coming but it is less clear how the
traditional interfaces with Robots, Programable logic controllers (PLCs Robots and Control Systems as
in Figure 2 work in an Enterprise Interoperability (EI) environment will interface with back office and
top office systems. Clearly new methodologies will need to be developed.
Figure 2: Key elements of knowledge support for future EI manufacturing systems
Focusing on the shop floor level, vendors such as Siemens, Rockwell, Omron, Schneider Electric
are rapidly deploying cloud-based solutions with Orchestration software to link shopfloor to IT systems
to Enterprise Resource Planning (ERP) systems. Bosch Rexroth is developing “Open Core
Engineering” [3] methodologies to allow systems integrators to choose the operating systems they wish
to deploy software directly into their control systems hardware.
These systems need interoperable middleware to “open up” or “connect” these islands of data to
provide significant opportunities to exploit the knowledge held within these systems. Further
enhancements mean that AI knowledge gained from the production line could directly influence
software steps that would otherwise be “hard coded” into a typical control systems program. So, while
the technology developments are allowing the direct manipulation of programmed steps, the method by
which these changes “on the fly” would be monitored or documented since AI systems are constantly
learning new algorithms and no two systems can ever be the same. So, if we take a typical line process
where a robot is picking and placing goods and then gets information to realise that the product
presented has minor defects and could be reworked, it needs to learn this and store it for future decisions.
4. Human-Robot Collaboration with Auditing System
The real challenge is to ensure that any changes made to the operation of a program carry an audit
trail because the likelihood of a system getting taught incorrect procedures needs to be reversible and
more importantly understandable to the control engineers that may need to override the machine
programs.
Figure 2: Presenting Data back down the Knowledge Chain
Maintenance staff need to understand how to re-commission a system if it has learned “bad habits”
which is inevitable if the system is self-learning.
Industreweb 4.0 (IW) [4] is a middleware software system that can take multiple data sources that
include AI learning systems that can learn to deal with human co-workers that need to interact safely
with the automation. IW takes these enhanced AI streams and combines it with standard control systems
software to not only keep the operator safe using a combination of vision systems and flexible program
manipulation. A key feature to track changes or “understand” the rational of the responses adopted by
the AI Control Systems is important to engineers who need to fix or repair processes that have failed or
are exhibiting irregular behaviour. This requires the ability to understand the rationale behind control
process decisions to correct or understand the reasons behind the failure. The rise of Explainable AI
(XAI) systems [5] is welcomed in the world of control systems as it allows engineers to continue to
interact, respond and maintain such systems keeping the decision making tasks firmly with the human.
5. Discussion and conclusions
AI Enhanced systems are likely to be making a big entrance to the shopfloor and take decision
making to the heart of the process. They will allow the current existing decision making processes to
focus on the business model and support the model by taking away the concerns of rescheduling
production tasks as they can be handled between the human operator and the support robot / cobot.
6. References
[1] https://en.wikipedia.org/wiki/Collective_intelligence
[2] Birkinshaw J. What Is the Value of Firms in an AI World?. In: Canals J., Heukamp F. (eds) The
Future of Management in an AI World. IESE Business Collection. 2020. Palgrave Macmillan,
Cham
[3] https://www.boschrexroth.com/en/xc/products/engineering/opencoreengineering/what-is-open-
core-engineering/index
[4] Industreweb Software developed by Control 2K Limited to act as the “Glue” to connect data
systems www.industreweb.com
[5] https://en.wikipedia.org/wiki/Explainable_artificial_intelligence