DNV Group Research & Development collaborates with nine universities across the world. This includes the sponsoring of professorships and PhD students, lectures by our employees and the supervision of students. One of the partnerships is with the Fudan University’s School of Data Science in Shanghai.
Collaborative research to test the boundaries of AI
Last year, the Artificial Intelligence Research Centre (AIRC) of DNV Research & Development started a research cooperation agreement with Fudan University in Shanghai, providing connection to wider industrial research and academic networks in China.
The collaboration with Fudan came only a year after DNV established the AIRC, led by Michael Chen, to benefit from working with partners in what is recognized as the leading AI research community in China.
AI is considered as a general-purpose technology that will have implications on every aspect of future DNV operations, as well as the company’s customers and society at large. DNV aims to develop new solutions based on AI technology, such as computer vision (whereby a computer can carry out tasks that require high levels of visual recognition) and smart industrial solutions, at the same time as providing future digital assurance technologies for systems enabled by the complex algorithms associated with AI.
For us at DNV, using AI for delivering our services in a more advanced way is relatively straight forward. But as certification service provider, we have the assurance of AI algorithms and AI-enabled systems in our research focus. Clearly, one day we want to certify AI. Therefore, we have established this collaboration and we are working with Professor Yanwei Fu at the School of Data Science of Fudan University, one of the most prestigious universities in China.
Dr. Ing. Pierre C. Sames
Group Research and Development Director
DNV
The School of Data Science was established by Fudan University five years ago in 2015. It is the first of its kind in China and is recognized for being the leader in research work in computer science. Highlighting its strengths, Computer Vision and Machine Learning Professor Yanwei Fu says, “Our research covers a wide range of interdisciplinary areas such as artificial intelligence, big data & analytics, digitalization, connectivity and cyber security to mention a few areas. Fudan University also has 35 departments that address science, engineering and medical disciplines.
“The collaboration with DNV started in June 2020 and will last for a minimum of three years. The field of collaboration is Assurance of AI and specifically focused on using what we call GANs (generative adversarial networks) to test AI algorithms. GANs are deep neural net architectures, and we are developing techniques of attack and defense in these deep neural networks,” adds Professor Yanwei Fu.
Pitting wits with artificial intelligence
Getting to the heart of the research Dr. Sames has some questions that need answers. How can AI be tested? How can we be certain that it works as expected? Can we get a warning if it is not? Can we open the ‘black box’ of AI?
Partially answering his own questions, Dr. Sames says, “Interestingly, we likely will use AI to test AI. As the task of probing an AI algorithm might be too complicated to perform, we consider using another AI algorithm as a testing tool. Our collaboration is focused on Deep Neural Networks and methods to test them and to make them more robust.”
The collaboration between DNV and the School of Data Science proposes an effective and universal method to defend against the adversarial examples generated from identified attacks and to explore the root causes of adversarial samples, explains AIRC managing director Michael Chen.
AI functionalities and components have been more and more used in assets and products across industries today. To provide assurance for these products, developing a systematic risk-based assessment method is a high priority. This method needs to be applied and validated in a real scenario, such as smart functionalities on an autonomous ship.
Professor Yanwei Fu agrees that the issue is a major one to be addressed.
AI is a very comprehensive topic and how to assure AI is a huge task. During our collaboration we have already made some progress. For example, we have proposed a bow-tie model approach which is based on concepts already in use in industrial applications by those that use the risk management model.
Professor Yanwei Fu
School of Data Science of Fudan University, China
The bow-tie method takes its name from the shape of the diagram used to identify causes and effects. It gives a visual summary of all plausible accident scenarios that could exist around a certain hazard and by identifying control measures, displays what can be done to control those scenarios. “So, by using AI-enabled systems we can make this approach even more advanced,” says Professor Yanwei Fu. “As an example of what we are working on with DNV, we focus on computer vision and IoT devices for use onboard autonomous ships. Deep learning is increasingly being used and there are also multiple traditional sensors which can be used as data sources. And cameras are also being added and are covered in our work to further advance AI assurance on vision capabilities.”
Playing tricks on the eyes of AI
One of the major issues with autonomous ships is how they will react to other vessels, fixed and floating objects in a safe and environmentally responsible manner. The ship will be receiving data from systems such as radar, AIS and GPS along with optical and perhaps Infrared images.
In a perfect world this would seem sufficient but at sea, faulty sensors, bad weather, poor quality data and even the images used in training the AI to identify what a ship looks like will affect the AI’s decision-making ability. This will be compounded by the fact that both the autonomous vessel and other ships will be constantly moving potentially allowing images to be misinterpreted.
There is a large body of published papers on this apparent vulnerability of AI in several applications and it will be something that needs to be overcome before the abilities of AI systems can be assured. The partners will therefore focus on using generative adversarial networks to test AI algorithms for testing detection and classification of ships.
“Our idea is to train a GAN on a set of customer-provided images, and then use it to generate new images that are realistic but slightly different from the training set. These new images would be used to check the robustness of a customer-created deep learning algorithm for detection and classification of ships. Eventually, such an approach will help us to build a toolbox for AI algorithm certification, complementing the approach described in DNVGL-RP-0510, “Framework for assurance of data-driven algorithms and models” says Michael Chen.
“Another example of our collaboration is the work we are doing on scenario modelling for risk-based assurance. Different standard levels will have different quantitative requirements. As we know, many hazard cases are due to human factors. One of the things we are aiming to do is to provide a standard library which manages knowledge and scenarios. The data-driven scene graphs will enable, for example, AIS analysis & modelling, AIS prediction & generation, scene classification and game engine simulations,” adds Professor Yanwei Fu.
According to Professor Yanwei Fu, the project is not yet at a stage where any findings or results can be published but he says this could happen soon and he is confident the work with DNV will produce beneficial results.
I believe our interdisciplinary research work and expertise provides a very good basis for collaboration with DNV, which is a recognized global company with a strong focus on helping its customers achieve safe and sustainable operations. Through our collaboration I think we can push AI assurance to a higher level and make a real difference to industries and societies,” he concludes.
