Innovationsleder på DTU
Det rette mindset er vigtigt for tech-startups, men penge i den tunge ende er også vigtigt. Når startups begynder at vokse, skal der være kapital til at vokse, uden at de skal ud af landet og til SIlicon Valley, Israel eller England, skriver Mark Riis.
Robotter genererer et hav af data, som man kan bruge til at optimere processer og forudsige slid. Forskere på Syddansk Universitet skal i et nyt projekt hjælpe Universal Robots med at udvikle AI-systemer, som kan være med til at forudsige, hvordan man optimerer programmeringen af robotterne.
Den Odense-baserede robotproducent Universal Robots har, siden de startede i 2008, solgt mere end 50.000 af deres samarbejdende robotarme, kaldet cobots. De har dermed skabt en hel industri for cobots, som giver virksomheder mulighed for at automatisere processer i produktionen, når der skrues, svejses, limes eller lignende.
En arbejdsstation, der styres af en robotarm, genererer et hav af data, som man kan bruge til at optimere processer eller til at forudsige slid på maskinen. Et samarbejde med Syddansk Universitet skal hjælpe Universal Robots med at udvikle it-værktøjer, der kan behandle de data, der er på robotten, og dermed udvide data med værdiskabende information for deres kunder.
“Der er rigtig mange data, så det er vigtigt at få gjort data tilgængelige på en brugbar og anvendelig måde for vores kunder. Det kræver, at du ved, hvad du har af muligheder, at det er nemt at anvende, og vigtigst af alt at det i sidste ende hjælper dig til bedre at løse de opgaver og udfordringer, der er vigtige for din virksomheds processer. Derfor er det vigtigt, at den information, der kommer ud, er af en sådan kvalitet, at man kan tage handling på den og dermed ændre sin adfærd,“ siger Morten Boris Højgaard, Innovation Lab Manager hos Universal Robots.
Kan optimere driften
De data, der bliver opsamlet, udgør et stort potentiale, som kan optimere den stabile drift af robotterne. Data kan nemlig give vigtig feedback på, om du bruger robotten forkert, og om der er risiko for, at robotten ikke kan udføre opgaven.
Det kan give anledning til nogle adfærdsændringer, så du som virksomhed lærer noget om din robot og i sidste ende også får mere ud af din robot. Det kræver bare, at der bliver lavet nogle gode løsninger, der er nemt tilgængelige og anvendelige for slutbrugeren.
“Det er et stort område, og der ligger noget forskning og udvikling i at komme frem til de gode værktøjer, som kan det her. SDU kan give os adgang til den nyeste viden, og de er eksperter i at designe de softwareløsninger. Omvendt kan vi bringe noget kontekst fra markedet,” forklarer Morten Boris Højgaard.
Skal afvikles tæt på robotten
Fra SDUs side er professor Mikkel Baun Kjærgaard særligt interesseret i at udvikle AI-systemer, som kan være med til at forudsige, hvordan man optimerer programmeringen af robotterne. Systemet skal designes, så det kan afvikles tæt på robotten, da robotterne som standard af sikkerhedsmæssige årsager ikke er koblet op på skyen. Det er afgørende for, hvordan man kan afvikle algoritmerne.
“Mere specifikt vil vi godt arbejde med, hvordan vi kan udnytte kunstig intelligens til at processere de her data, så vi kan lave forudsigelser og vedligehold. Vi skal samtidig kunne afvikle det tæt på udstyret, da der kan være nogle begrænsninger i forhold til datasikkerhed, der gør, at man ikke kan sende data op i skyen. Det kræver nogle andre softwareteknologier kontra den normale model, hvor man sender det op i skyen,” forklarer professor Mikkel Baun Kjærgaard fra Mærsk Mc-Kinney Møller Instituttet, Syddansk Universitet.
Samarbejde kan give forspring
Samarbejdet mellem Universal Robots og Syddansk Universitet er det første “Bridge-projekt”, som foregår i regi af DIREC, der er et nyt partnerskab mellem syv universiteter og Alexandra Instituttet. Det er støttet af Innovationsfonden og har til formål at udbygge kapaciteten inden for forskning og uddannelse i digitale teknologier i Danmark.
Dette første projekt er et godt eksempel på, at universiteter og virksomheder har forskellige former for ekspertise, som de kan drage fælles nytte af, lyder det fra Thomas Riisgaard Hansen, der er managing director for DIREC.
“Du kan godt have en hypotese om, at en algoritme kan udføre en bestemt opgave, men reelt set ved du det ikke som virksomhed. Men det er der sikkert nogle forskere, som har arbejdet med. På den måde kan du få et enormt forspring som virksomhed, fordi du ikke skal lave alle eksperimenterne selv. Omvendt har man som forsker også behov for at få kendskab til, hvad det egentlig er for nogle problemstillinger, som man har i samfundet. Det giver en naturlig prioritering af, hvad der er interessant at arbejde med, ” forklarer han.
Fakta om Bridge-projektet:
Projektet er det første “Bridge-projekt”, der foregår i regi af DIREC, der er et nyt partnerskab mellem syv universiteter og Alexandra Instituttet, som skal udbygge kapaciteten inden for forskning og uddannelse i digitale teknologier.
I projektet arbejder man med forskellige cases og på tværs af universiteterne i forhold til udvikling af teknologier. Ud over samarbejdet mellem Syddansk Universitet og Universal Robots, så arbejder IT-Universitetet i København på cases med Novo Nordisk og Energinet. DTU arbejder på en case med virksomheden Octavic.
I projektet kigger man på alt det udstyr, som hjælper os med at holde vores energi- og industriproduktion i gang. Man vil gerne have, at det kører så stabilt som muligt. Her vil man gerne udnytte de data, som man indsamler, til at foretage bedre vedligehold af udstyret, der bygger på forudsigelser, således at det kører mere stabilt.
Fakta om Universal Robots (UR)
Universal Robots har som mål at forbedre arbejdsformer gennem sin førende robot-platform. Siden introduktionen af den første kollaborative robot, også kendt under betegnelsen cobot, i 2008, har Universal Robots udviklet en bred portefølje med modellerne UR3e, UR5e, UR10e og UR16e med varierende rækkevidde og løfteevne. Hver model er understøttet med et bredt udvalg af tilbehør og software i UR+økosystemet. Dermed kan UR cobots anvendes i en bred vifte af industrier og på fleksibel vis tilpasses forskellige opgaver. Universal Robots, som er ejet af amerikanske Teradyne inc. har hovedsæde i Odense og kontorer i en lang række lande inklusiv USA, Tyskland, Frankrig, og Indien. Universal Robots har solgt over 50.000 cobots globalt. Læs mere på www.universal-robots.com
Yderligere info:
Morten Boris Højgaard, Innovation Lab Manager hos Universal Robots, +45 51 77 73 54, mbho@universal-robots.com
Mikkel Baun Kjærgaard, professor ved Mærsk Mc-Kinney Møller Instituttet, Syddansk Universitet, +45 21 97 24 47, mbkj@mmmi.sdu.dk
Thomas Riisgaard Hansen, managing director for DIREC, +45 29 40 33 97, thomas.r.hansen@direc.dk
7 September 2021
Explore projects are small agile research and innovation projects with the purpose to quickly screen new ideas within or between the core thematic topics of DIREC – possibly in relation to specific challenges of companies or society. Explore projects run for 3-12 months with a focus on identifying and creating new research challenges and areas.
Projects can be stand-alone projects or part of a sequential evolution of projects. For example, an Explore project may be a natural start to investigate into a new field or topic, which can lead to the creation of a larger research project.
DIREC is launching two special calls for Explore project proposals:
Each Explore project may be supported with up to DKK 300-500.000 including overhead for a period of up to a year.
Deadline for applying is November 5th, 2021.
Looking forward to seeing your proposals.
We will discuss the needs of industry, current solutions and the scientific and technical challenges that are connected to the problem of an efficient re-use of robot data.
However, the potential of re-using robot data is not realized yet due to scientific, technical and IPR issues. The ReRoPro project, Re-Use of Robotic-data in Production through search, simulation and learning aims at addressing this problem.
We invite for tasks such as predictive maintenance, speeding-up the establishment of new assembly solutions and fine-tuning of critical components of your production.
At the workshop, we will discuss the needs of industry, current solutions and the scientific and technical challenges that are connected to the problem of an efficient re-use of robot data.
Find out how to make efficient use of your robot data!
9:30 Welcome with coffee and rolls in the lobby
10:00 Welcome and Introduction
Prof. Norbert Krüger, University of Southern Denmark, coordinator of the ReRoPro-project
10:15 Session 1: The ReRoPro Project
10:15 The i4.0 use case and current software solution
Assoc. Prof. Aljaz Kramberger, University of Southern Denmark
10:30 Data sensitivity issues in data re-use
Prof. Mikkel Baun Kjærgaard, University of Southern Denmark
10:45 Differential simulators, digital twins & data (d^3)
Prof. Kenny Erleben, University of Copenhagen
11:00 Data-reuse in machine learning and planning
Assoc. Professor Andres Masegosa, Aalborg University and Assoc. Professor Alvaro Torralba, Aalborg University
11:15 Coffee Break
11:30 Session 2: Data-Re-Use in Industry and Science
11:30 Examples of industrial robot applications, where
data could have been re-used
Prof. Henrik G. Petersen, University of Southern Denmark, MADE
11:50 The Danish robot ecosystem and the re-use of data
Project Manager Søren Adamsen Mouritzen, Odense Robotics
12:05 Paving the way for a Cambrian explosion in robotics: open knowledge services for robotics applications
Prof. Michael Beetz, University of Bremen
12:45 Lunchbreak
13:30 Session 3: Industrial perspectives
13:30 Data in Pharmaceutical Device Manufacturing
Senior Assembly Engineer, Roger de Reus, Novo Nordisk
13:45 Robots and Sensors in Surface Treatment Applications
Head of R&D Lars Kristian Feddersen, Nordbo Robotics
14:00 Re-use of data in CAM programming
Head of Production Bo Schmidt, WellTec and Head of Digital Integration CNC Thomas Hyllen, WellTec
14:15 Factory of the Future
Program Director Alex Severin, Rockwool
14:30 So we got big data – now what?
Chief Commercial Officer Anders Meister, CIM.AS
14:50 Coffee Break
15:20 Toward a Knowledge-Based Data Backbone for Seamless Digital Engineering in Smart Factories
Dr. Markus Rickert, Technische Universität München
15.50 Round Table discussion
16:30 Concluding remarks
The workshop is organized by DIREC with MADE as a supporting partner and with support from DDSA (Danish Data Science Academy.
The event is held in collaboration with IDA Fremtidsteknologi. The participant list will be shared with IDA for statistical use only.
31. maj 2022
Ny database skal samle roboterfaringer fra industrien i skyen, så virksomhederne ikke behøver at starte forfra hver gang, de skal indsætte robotter i produktionen.
Legal principles, philosophy and doctrines are the pillars of modern society. It is tempting to believe that, while specific laws and regulations adapt to the particular technologies of the time, the basic legal doctrines remain unchanged – and guide us in regulating and harnessing technology.
This talk will present situations where technological feasibility, accompanied with appropriate theory-of-computation reasoning, impacts not only specific laws and regulations, but also some basic legal doctrines. Specifically, these are situations where justified secrecy and privacy rights regarding sensitive information is pitted against equally justified transparency, accountability, and due process rights pertaining to the same information.
Current legal doctrines accept the seemingly inevitable reality the not all rights can be honored, and instead aim at “balancing the harms” on all sides. In sharp contrast, cryptographic concepts such as Zero-Knowledge Proofs and Secure Multi-Party Computation enable legal processes that do honor all rights.
Moreover, these technologies enable fine-grained delineation of what partial information should be disclosed and to whom, thus opening the door to more nimble legal doctrines and thinking regarding the ownership, sharing, and use of information in a modern society.
The talk will be mostly based on the following two papers:
Professor of Computer Science,
Boston University
Ran Canetti is a professor of Computer Science in Boston University, where he directs the center for Reliable Information System and Cyber Security. He is a Fellow of the Association for Computing Machinery and the International Association for Cryptologic Research, and an incumbent of the RSA Award in Mathematics.
Canetti’s research interests lie primarily in cryptography and information security, with emphasis on the design, analysis and use of cryptographic algorithms and protocols. Recently he has been studying ways for the co-design of algorithms, law, and policy so as to provide sound foundations for society in the information age.
24. august 2021
Klumme: Det skal være nemmere for virksomheder at få adgang til it-forskning og it-kompetencer. Det er absolut nødvendigt, at vi styrker rammevilkårene for, at digitale teknologier samt viden og kompetencer om disse teknologier kommer flest mulige virksomheder til gode.
Nye tal fra EU viser, at 50 procent af nye vækstvirksomheder har en base i it og/eller digitale teknologier.
Samtidig viser undersøgelsen ”Digital competitiveness ranking 2020”, udgivet af World Competitiveness Center IMD, som er en sammenligning af 64 landes evne til at anvende digitale teknologier som en nøgledriver til økonomisk transformation i erhvervslivet, den offentlige administration og det bredere samfund, at Danmark er med helt fremme med en flot tredjeplads, kun overgået af USA og Singapore.
Tidligere ph.d.-studerende på IT-Universitetet, Javier Gonzalez, blev ansat til at være med til at etablere Samsungs Memory Solutions første Research and Development-center i Europa.
Valget faldt på tidligere ph.d.-studerende på IT-Universitetet, Javier Gonzalez, da Samsung Memory Solutions manglede medarbejdere, som ud over at mestre Linux til fingerspidserne også kunne opbygge et open source økosystem, som gør det nemmere for kunderne at bruge Samsungs produkter.
11 August 2021
Den 6.-9. august mødes de 16 bedste hold fra hele verden for at konkurrere i det uofficielle verdensmesterskab i hacking. Hackingkonkurrencen, kaldet DEFCON, afholdes årligt i Las Vegas, men på grund af COVID-19 afholdes den delvist online. Det danske hold “Kalmarunionen” deltager i et nordisk drømmehold med spillere fra Sverige og Norge.
From August 1 to August 4, 2022, the Departments of Computer Science at ITU Copenhagen and Aarhus University invite you to the Summer School on Privacy-Preserving Machine Learning.
Privacy-Preserving Machine Learning is an important and exciting research subject that investigates how to benefit from machine learning techniques while preserving the privacy of training data and learned models.
At the PPML School 2022 lecturers with both a theoretical and applied background will cover a broad spectrum of subjects such as Multiparty Computation, Fully Homomorphic Encryption, Differential Privacy, Federated Learning as well as practical attacks. Current confirmed speakers are:
The school is aimed at PhD and Master students in the areas of Security as well as Machine Learning, but we also encourage researchers as well as other people with an interest in the area to attend.
Registration for the school is now open for a fee of 500 DKK (approximately 70 USD or 67 EUR). Students can obtain 3 ECTS for attending the school.
The event is organized by Bernardo David, Associate Professor at ITU Copenhagen and Carsten Baum, Assistant Professor at Aarhus University and will take place from August 1st until August 4th on the campus of ITU Copenhagen. We are currently investigating a remote participation option, but this is so far not decided.
More information will be provided soon. We will provide information about potential stipends at a later point of time.
Registration deadline is on June 30th!
The event is supported by the International Association for Cryptologic Research, the Danish Data Science Academy, the Pioneer Centre for Artificial Intelligence as well as the Digital Research Centre Denmark (DIREC).
17 JUNE 2022
Automated systems like self-driving cars and AI-based decision support are becoming an increasingly large part of our everyday lives, and so is the need for modelling and verification of the software running these systems. At the MOVEP 2022 Summer School, hosted by the Department of Computer Science, Aalborg University, leading researchers, students and people from the industry convened to discuss challenges and opportunities within this field.
By Stig Andersen, Aalborg University
The five-day MOVEP Summer School 2022 (June 13-17) on modelling and verification of parallel processes had attracted 70+ participants, primarily PhD students, but also people from the industry.
With the lecture hall of the Department of Architecture, Design and Media Technology right at Aalborg’s harbour front as a great venue, they enjoyed a packed programme of talks and tutorials from 11 leading researchers on model checking, controller synthesis, software verification, temporal logics, real-time and hybrid systems, stochastic systems, security, run-time verification, etc.
One of the speakers was Christel Baier, Professor and Head of the chair for Algebraic and Logic Foundations of Computer Science at the Faculty of Computer Science of the Technische Universität Dresden, and together with Joost-Pieter Katoen, the author of a key publication in the field, Principles of Model Checking (MIT Press, 2008). She has been working within the broad field of verification and analysis techniques for stochastic operational models for more than twenty years.
– I really had not expected to work so long within this area, but as it often turns out in science, apparently simple problems are not at all simple and will require more research. So, if the students at this summer school would take the message that this is an exciting and very important field and choose to explore it further, I would be very happy. MOVEP is a very nice event, and being able to come to Denmark and not least being able to meet again after the Corona shutdown is really great, she says.
Another speaker was Nir Piterman, Professor in the Department of Computer Science and Engineering, University of Gothenburg and Chalmers, and a prominent figure within formal verification and automata theory. He kicked off the summer school programme Monday morning with a tutorial on reactive synthesis, which is a technique for automatically generating correct-by-construction reactive systems from high-level descriptions.
– In my tutorial, I tried to give the participants a taste of the so-called discrete two-player turn-based games technique, where you think about the environment as one player and the system as another player. The interaction is like a game between the two, and the system has to come up with a strategy to satisfy some goal, he explains.
Nir Piterman also sees an event like MOVEP as a very good opportunity for young researchers to be exposed to concepts and techniques that they would not necessarily be exposed to otherwise.
– It is my hope that the talks and tutorials at this event will fertilize their work and provide them with new ideas about how to apply these techniques in different fields. One possible usage of two-player games is synthesis, but the usage could be wider and potentially applied to other problems, he says.
Nir Piterman is currently the holder of an ERC consolidator grant to study the usage of reactive synthesis for multiple collaborating programs.
In her tutorial, Christel Baier focused on explication, which refers to a mathematical concept that in some way sheds light on why a verification process has returned a given result.
– Explainability is important. We have to make systems more understandable to everyone – scientists, designers, users, etc. Today, everybody is an IT user, so this is not only relevant for computer scientists, she says.
According to Christel Baier, there is a higher purpose:
– Since systems make decisions, users should have the opportunity to understand why decisions were made. Moreover, users should be supported in making decisions by themselves and be given an understanding of the configuration of these systems and their possible effects. Again, it comes down to the question of cause and effect, which was a recurring theme of my tutorial.
The research on the results presented by Christel Baier at her tutorial has been carried out within and is motivated by the missions of the collaborative projects “Center for Perspicuous Computing (CPEC)” and “Centre for Tactile Internet with Human-in-the-Loop (CeTI)”.
Research within modelling and verification of parallel processes may also explore the question: Could we automatically generate systems that perform exactly according to the specifications instead of checking afterwards that they do? Nir Piterman dealt with this topic in his tutorial.
– Techniques to automatically generate correct-by-construction reactive systems from high-level descriptions have been explored in academia for quite a number of years. It has proven to work in some domains, but it would not be realistic to set as an ambition to build one synthesizer that you feed a specification to and expect it to auto-generate safe and error-free systems for all possible programming domains, he says.
According to Nir Piterman, the most successful applications so far have been within robotics. However, this success makes us think about what is the meaning of correct-by-construction.
– What does “correct” really mean? If it means that the system does exactly what was described in the specification, what happens if the specification is flawed? So, the focus of the correctness problem might change: Rather than making sure that the system matches the specification, the task is to ensure that the specification is thorough enough and reflects what the designer had in mind.
FURTHER INFORMATION
CONTACT
Martin Zimmermann
Associate Professor
Department of Computer Science
Aalborg University
Mail: mzi@cs.aau.dk
Phone: +45 9940 8770
Stig Andersen
Communications Officer
Department of Computer Science
Aalborg University
Mail: stan@cs.aau.dk
Phone: +45 4019 7682
Professor Nir Piterman, University of Gothenburg and Chalmers
Professor Christel Baier, Technische Universität Dresden
The aim of DIREC is to expand the capacity within research, innovation and education in digital technologies in Denmark. In addition, DIREC shall contribute to the competitiveness of Denmark through collaboration with Danish businesses and the public sector on developing new innovative products and services based on the newest digital technologies.
DIREC is partially funded by Innovation Fund Denmark.
© 2021 All Rights Reserved