KICK

Artificial Intelligence for Campus-Communication

News

Slide 1
KICK White Paper

Leveraging Machine Learning for Industrial Wireless Communications
Two main trends characterize today’s communication landscape and are finding their way into industrial facilities: the rollout of 5G with its distinct support for vertical industries and the increasing success of machine learning (ML). The combination of those two technologies open the doors to many exciting industrial applications more

Slide 1

Bosch puts first 5G campus network into operation

5G to be deployed in Bosch plants worldwide
Bosch is putting its first 5G campus network into operation. At its Industry 4.0 lead plant in Stuttgart-Feuerbach, the company aims to manufacture under previously unheard-of conditions, with data being transferred extremely reliably and ultra-fast, and machines reacting almost instantaneously. For the first time, wireless implementation will be possible for critical applications that require absolute precision and safety. Without exception, people and machines will be able to cooperate safely and without barriers. “5G strengthens our competitiveness and lets us make even more of Industry 4.0’s potential,” says Dr. Michael Bolle, board of management member and CDO/CTO at Bosch. “We will gradually roll 5G out to our roughly 250 plants around the world.” The locations where Bosch will be setting up 5G networks in the coming months include its research campus in Renningen, Germany. The company is also developing 5G-capable products and launching its first solutions for industrial use.
more

previous arrowprevious arrow
next arrownext arrow

Project Description

Key Figures

The project Artificial Intelligence for Campus-Communication (KICK) is a research project funded by the Federal Ministry for Education and Research (BMBF), from the announcement Artificial Intelligence in Communication Networks, which researches the applicability of Artificial Intelligence (AI) in future private and public 5G campus networks.

Total Cost

0

mio Euro

Total funding

0

mio Euro

Funding rate

0
%

Month

0
01.2020 - 12.2022

Project Partner

0
Industry & Research

Project Objective

The goal of KICK is to significantly simplify and improve the operation of future 5G campus networks by using AI methods. The focus here is on Industry 4.0 environments with their high reliability and latency requirements. Specifically, an AI framework for the tactical and operational management and control of such communication networks is being developed and investigated, taking into account the limited communication and computing resources. On the one hand, the research work includes the definition of requirements for AI algorithms and the identification of suitable data and data formats. On the other hand, questions concerning the training, adaptation, compression, exchange and interaction of AI algorithms are addressed. In order to meet the high requirements of industrial applications, hybrid, i.e. data- and model-based approaches are pursued and data from production is linked with data from communication networks. Transfer Learning based on AI network state detection is used to derive adequate behaviour also on the basis of similar situations or similar logical network instances. This enables robust, continuous configuration, optimization and error handling, thus fully exploiting the potential of campus networks and networked Industry 4.0 systems. The project is based on experimental work with real communication and production data from a real factory environment. The automation advantages achieved are also demonstrated in an exemplary manner in such an environment.

User Stories

KICK AG, a leading producer of intelligent and autonomous machine tools has a modern production sites and is very successful in their market domain. All sites belonging to KICK AG are equipped with the latest automation devices, a modern industrial Ethernet based on TSC and a private 5G campus network. Everything works smoothly with high efficiency. Much work for operation is done by different groups, e.g., the campus network is managed and operated by an external company, the devices and SW functions in the factory have external support or may be leased products, i.e., in a pay-as-you-use fashion and several applications use in-device or edge-cloud and external cloud-based services.

Production Line Upgrade

To be able to fulfill the high demand and extend the production capacity and flexibility, the management decides to acquire a less modern factory hall which needs to be updated to fulfill all the company standards. In the course of site modernization, the following steps are planned. First, 3D maps of both operating and empty factory halls are created. Afterwards, the network planning can be carried out smoothly, thanks to the innovative AR enabled 3D network planning tool, where in real time a map of the campus is overlapped to the network deployment, showing propagation condition and specific KPIs distribution, such as a reliability map of the factory hall. This way, the best access point locations are automatically detected. Once the network is deployed, the service accessibility in each planned machine location is verified and supported by the creation of a digital twin of the factory line. Finally, old existing machines are retrofitted by means of digitalization and connection to 5G network whenever possible.

After performing the upgrade, it is planned to order and install a new large device, namely a 3D printer, which requires changing also the position of nearby machines as well as guaranteeing connection to those. There is also a risk that the device – due to its size – changes radio reception conditions in the surroundings.

As a first step, the device must be integrated into the IT infrastructure (including network), the production and internal logistics processes. Due to its large size, the introduction of the new device changes the wireless propagation properties of the factory hall. On the other hand, the frequent interaction with other devices in terms of both information exchange and material flow changes the network traffic pattern. The 5G campus solution provided by KICK automatically recognizes the changes and adapts to the new situation. In summary, different from former days, where the installation of new large devices and machines as well as changes in the production line would have caused long periods of interruption in the production processes, thanks to the innovative solutions, such interruption times can be shortened to a fast and easy re-planning and re-configuration of the network.

Finally, the factory owner can control operations using an AI/ML based analysis tool which monitors data/logs in real time to ensure that the system works as expected and all the external parties comply to security rules and provide the service in the expected quality. After putting the device into service, the AI-based monitoring system proposes post-optimizations, warns for potential pitfalls, anomalies and visualizes the system health.

In summary, the truly plug&work solution will provide the following benefits:

  • Virtual commissioning allows to “install and test” changes in the factory hall, e.g. deployment of a new device, in the digital twin of the factory.
  • New devices are automatically detected and connected to the enterprise network via 5G (or Time Sensitive Network) and are smoothly integrated into the campus network.
  • Thanks to the AI-based radio management, the 5G network is adapted on the fly without noticeable impact on the adjacent devices.
  • In case physical changes to the network are needed, such as installing or moving an access point, or changing the configurations of the access point, an augmented reality tool is available to guide the workers into the process.
  • Intelligent middleware components integrate the new device into the production and make the onboarding as easy as possible.
  • The integration process can be monitored and controlled over an easy-to-use graphical user interface.

Intralogistics and Mobile Robotics

Peter Production Planner is working on the digitization of the factory floor to improve the product throughput and to reduce the number of customer complaints. Any ordered good is already tracked before it arrives at the KICK AG factory hall. To ensure a smooth unloading of trucks, automated guided vehicles (AGVs) are pre-scheduled and informed about the routes before the truck arrives. In addition, the communication network is configured such that the new devices can be seamlessly integrated, while still keeping security constraints such as isolation by means of network slicing. To ensure faultless goods, an automated scanning of the boxes with the goods is initiated upon arrival. The AGVs automatically load the goods and follow their path from the factory entrance to the machines for further processing.

Peter Production Planner must guarantee that AGVs arrive safely and fast at the machines. Therefore, he ensures that the navigation is not influenced by shadowed areas and that the maps (factory floor and radio maps), a kind of digital twin of the factory, are constantly updated. With these real-time maps, anomalies, such as a dropped load can be detected and also broken AGV can be identified and maintained/serviced remotely. In case of such an incident, other AGVs are automatically redirected. As the weight of the goods might exceed the maximum cargo capacity of an AGV, a virtual compound/platoon of AGVs is formed, that lets the AGVs cooperatively navigate through the factory. Furthermore, the AGVs are constantly monitoring for alarms, in which case they vacate the premises in a pre-planned manner to allow for unobstructed operations for the first responders.

To ensure a smooth transition from the AGV to the machines, where the goods are processed, a communication link between the goods and the machines is set up and the machines get their processing instructions. During the processing of the goods, the tags of the different components are combined so that the final product only has a single tag for identification and communication. In case of a machine outage, the production process is automatically rescheduled (to a spare machine, other processing step) and the AGVs are redirected accordingly.

Before the products are picked up by the AGVs, an automated, offloaded quality control ensures compliance with the high-quality standards of Peter and the KICK AG. The AGVs transport the products from the machines to an intermediate storage, where they are packed and stored. During the navigation of the AGVs, they are constantly monitored, which helps Peter Production Planner to predict AGV outage and consequently increase the efficiency/rentability of his production process. In addition, the AGVs also help to detect anomalies caused by jamming or some other unusual interference. A final security check ensures that the products and their surrounding boxes are not damaged before they are loaded by the AGVs onto the trucks.

In summary, the efficient intralogistics and mobile robotics solution will provide the following benefits:

  • Speed up of production process due to increased automation
  • Identification and report of white spots in the factory hall
  • Automatic network adjustment to ensure high QoS for AGVs
  • Automatic update of digital twin as well as anomaly detection based on continuous network measurements of the AGVs
  • Product tracking throughout the entire production process.

Data Analysis and Predictive Maintenance

The KICK AG successfully implemented the industry 4.0 in their smart factory. A part of their success lies in the fact that, contrary to the competitors, they are able to handle the enormous amount of data generated by the individual intelligent machine, ERP-system and 5G campus network.

The KICK AG  is not only collecting data, but also continuously analyze it with the help of AI algorithms to gather several information from the single raw data source such as anomaly detection. Thus, in the case of detected anomalies, the AI system lists the root causes of the anomaly and the employee in the smart factory receives an alarm that an anomaly has been identified. At the same time to avoid any damage to the machine, the system degrades the machines automatically to “limited operation”. The employee goes to the UI and looks for the causes and the expected impact of the identified anomaly and checks if any countermeasures are already available for that. If there is no any solution defined against the identified anomaly and no expert  is available at the site, the employee uses his AR glasses for the help of the remote expert to fix the problem. Since there is a strong interconnection between the network management and the production environment, and the network management is also aware of the problem, they provide the required network  resources for the AR usage. The help of the remote expert enables a smooth and fast resumption of machine operation without any significant production downtime. After resolving the anomaly with the remote expert, the solution is added in the available solutions pool against the identified anomaly. In this way the solution database also gets stronger in the course of time.

The benefits of data analysis and predictive maintenance can be summarized as below

  • Automated anomaly detection
  • No or minimal downtime by limiting the functionality instead of complete stop
  • Faster resolution of the problem either from existing solution or remote support

Work packages

1

Coordination and technical management of the project. Coordination is done by Nokia-S and technical management is done by HHI.

2

Definition of use cases for industrial, non-public networks and their analysis with regard to requirements for the AI. Selection of suitable AI functionalities, which are optimized in AP3 and AP4 for the use cases. Interface definition for data extraction from non-public communication networks within real production environments as well as production data for generating context information are also part of the AP. Development of system models for a cyber-physical network simulator to provide synthetic training data.

3

Development and performance evaluation of AI-based algorithms for use in planned changes in communication networks and production environments. Recording and description of the actual state of complex industrial communication systems including relevant real world influences on the basis of single measuring points which are to serve as a basis for later optimization. The optimization shall benefit in particular from already existing knowledge. Thus, an AI-supported evaluation of the effects of an optimization measure is to be made, sources of error are to be identified and knowledge from previous reconfigurations is to be transferred to new situations by means of transfer learning. A challenge in the industrial environment is the integration of expert knowledge from the fields of communication and production as well as different data sources and models into a robust overall system. AP3 uses simulations developed in AP2 as well as methods and interfaces for the collection of measurement data. Selected AI methods are demonstrated in AP5 with prototype systems.

4

Development of management and orchestration functions or algorithms that automate the operational network management of campus networks using AI-based methods The focus is on the special features and use cases of operational campus network management. Frequent reconfigurations of the entire infrastructure must be taken into account, which not only influence tactical network management but also operational management functions. These must be particularly robust against changes in context. The AI models must be robust against changes and should be adapted more quickly to changes using "transfer learning" methods. 

5

Collection of measurement data in application scenarios as real as possible for training and evaluation of the AI procedures to be developed in WP 3 and 4 Realisation of selected application cases in a real environment in the sense of a proof of concept, in which different aspects of potentially different partners (e.g. different AI procedures) interlock. Qualitative and - as far as possible - quantitative evaluation of the achievable improvements compared to the state of the art.

6

To link measures for the publicity of the project and results and for KICK profitably with suitable projects and committees, e.g. in connection with standardisation activities. Publications, panel(s), workshop(s), demo(s) or posters at relevant flagship conferences such as IEEE Globecom, ICC, NOMS/IM, NIPS etc. are aimed for. Such activities will not be funded, but will be pursued as they are in the interest of the consortium.

Publications

  • Malanchini, Ilaria, Patrick Agostini, Khurshid Alam, Michael Baumgart, Martin Kasparick, Qi Liao, Fabian Lipp, et al. “Leveraging Machine Learning for Industrial Wireless Communications.” ArXiv:2105.02051 [Cs, Eess], May 5, 2021. http://arxiv.org/abs/2105.02051.
  • Jiang, Wei, and Hans D. Schotten. “Recurrent Neural Networks with Long Short-Term Memory for Fading Channel Prediction.” In 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), 1–5, 2020. https://doi.org/10.1109/VTC2020-Spring48590.2020.9128426.
  • Jiang, Wei, and Hans D. Schotten. “A Deep Learning Method to Predict Fading Channel in Multi-Antenna Systems.” In 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), 1–5, 2020. https://doi.org/10.1109/VTC2020-Spring48590.2020.9129077.
  • Jiang, Wei, and Hans Dieter Schotten. “Deep Learning for Fading Channel Prediction.” IEEE Open Journal of the Communications Society 1 (2020): 320–32. https://doi.org/10.1109/OJCOMS.2020.2982513.
  • Agostini, Patrick, Zoran Utkovski, and Slawomir Stanczak. “Not-Too-Deep (N2D) Channel Charting,” 2021.
  • Komuro, K., M. Yukawa, and R. L. G. Cavalcante. “Distributed Sparse Optimization with Minimax Concave Regularization.” Proc. IEEE Statistical Signal Processing Workshop, July 2021.
  • Hu, Tianlun, and Qi Liao. “Real-Time Camera Localization with Deep Learning and Sensor Fusion.” In ICC 2021 - IEEE International Conference on Communications, 1–7, 2021. https://doi.org/10.1109/ICC42927.2021.9500770.

Contact

Dr. Ilaria Malanchini
Senior Research Engineer
End-to-End Network & Service Automation
Nokia Bell Labs

Mobile: +49 175 7266815
E-mail: Ilaria.Malanchini (at) nokia-bell-labs.com

Nokia Solutions and Networks GmbH & Co. KG
Lorenzstr. 10, 70435 Stuttgart
Sitz der Gesellschaft: München / Registered office: Munich
Registergericht: München / Commercial registry: Munich, HRA 88537
WEEE-Reg.-Nr.: DE 52984304

Disclaimer

Dr. Ilaria Malanchini
Senior Research Engineer
End-to-End Network & Service Automation
Nokia Bell Labs

Mobile: +49 175 7266815
E-mail: Ilaria.Malanchini (at) nokia-bell-labs.com

Nokia Solutions and Networks GmbH & Co. KG
Lorenzstr. 10, 70435 Stuttgart
Sitz der Gesellschaft: München / Registered office: Munich
Registergericht: München / Commercial registry: Munich, HRA 88537
WEEE-Reg.-Nr.: DE 52984304
Persönlich haftende Gesellschafterin / General Partner: Nokia Solutions and Networks Management GmbH
Geschäftsleitung / Board of Directors: Dr. Wolfgang Hackenberg, Ralf Niederberger
Vorsitzender des Aufsichtsrats / Chairman of supervisory board: Hans-Jürgen Bill
Sitz der Gesellschaft: München / Registered office: Munich
Registergericht: München / Commercial registry: Munich, HRB 163416

The contents and works on these pages created by the site operators are subject to German copyright law. The reproduction, editing, distribution and any kind of use outside the limits of copyright law require the written consent of the respective author or creator.

External links and disclaimer The references made by links to the contents of external Internet sites are for information purposes only. The responsibility for these external contents lies solely with the provider who provides the contents; the contents are not adopted as his own. We assume no liability for any data protection or other legal infringements in other offers to which we have set a link.

The protection of your personal data is a basic right. The KICK consortium will process your personal data only in good faith for specified purposes. In this privacy policy we inform you about the handling of your personal data when visiting and using our website.

The visit and use of our website is basically possible without the collection of personal data. Personal data is only collected on our website to the extent technically necessary. The collected data will not be passed on to third parties, except in cases of specific legal obligations.

We would like to point out that data transmission on the Internet when visiting and using our website is protected and therefore cannot be easily eavesdropped on or falsified by third parties.

Every time our website is accessed, usage, traffic and connection data is collected and temporarily stored in a log file.

The following data is stored for each access/retrieval:

  • IP address
  • transmitted user-agent information (in particular type/version of web browser, operating system, etc.)
  • transmitted referrer information (URL of the referring page)
  • Date and time of access/retrieval
  • Transmitted access method/function
  • transmitted input values (search terms, etc.)
  • the page or file called up
  • data volume transferred
  • Status of access/retrieval processing

The stored data records can be statistically evaluated in order to adapt and optimise our website to the needs of our visitors. Techniques that make it possible to trace the access behaviour of the users are not used. The creation of user profiles (tracking) is excluded.

Die gespeicherten Datensätze sind nicht bestimmten Personen zuordenbar. Eine Zusammenführung mit anderen Datenquellen wird grundsätzlich nicht vorgenommen. Wir behalten uns jedoch vor, die gespeicherten Daten nachträglich zu prüfen, wenn uns konkrete Anhaltspunkte für eine rechtswidrige Benutzung bekannt werden.

On our website will use so-called cookies. Cookies are small files that are stored by your web browser. The cookies used on our website do not cause any damage to your computer and do not contain any malware. They enable a user-friendly and effective use of our website.

We transmit so-called session cookies to your web browser. They are only valid for the duration of your visit to our website and have no meaning outside our website. The session cookies are needed to identify your session during your visit with a unique number and to transmit our content in your preferred language. At the end of your visit, the session cookies are automatically deleted when you close your web browser.

We also transmit permanent cookies to your web browser with a maximum validity of 365 days. We use these cookies exclusively to observe your settings for the type of display (normal, inverted) and for the font size the next time you visit our website. Furthermore, we will note whether you have taken note of the information on the use of cookies in your web browser.

You can set your web browser to inform you about the setting of cookies and to allow cookies only in individual cases or to exclude the acceptance of cookies for certain cases or generally. You can also set your web browser to automatically delete cookies when you close it. If you deactivate cookies, the functionality of our website may be limited. In any case, our information offer is available in its entirety.

We do not integrate social media plugins into our website. Therefore, when you visit our website, no data is transmitted to social media services. A profile creation by third parties is therefore excluded. However, you have the option of switching directly to our information services on Facebook, Twitter and YouTube. For the use of these services we refer to the data protection regulations of the respective service providers.

You have the possibility to contact us by e-mail. We use your e-mail address and other personal contact data exclusively for correspondence with you. Due to legal obligations, all e-mail correspondence is archived. If necessary to protect our legitimate interests, your e-mail address and other personal contact data from correspondence may be stored in our contact database.

en_USEnglish