SI.20.002 – Smart Industrial Evolution – Next-generation Real-World Problem Solving for Industry

Route: Smart industry

Cluster question: 054 How do we ensure that the Dutch economy remains competitive?

In the age of AI, algorithms are omnipresent, and quality of life depends increasingly on quality of algorithms. This implies a major challenge: enabling non-experts to solve the broadest possible range of problems with the best possible algorithms. Creating efficient and effective algorithms requires experts with much AI knowledge. Such manual craftsmanship does not scale, and the worldwide shortage of AI experts already limits the competitiveness of many businesses. This creates a strong demand for easy-to-use, yet powerful, optimization technology to be able to remain competitive in the global economy. This project offers a solution: Automated Evolutionary Computation – AutoEC, based on a novel combination of evolutionary computation and machine learning, both cornerstones of AI. Evolutionary algorithms (EAs) are the right basis because, in the hands of experts, they have proven to offer large practical advantages in solving complex real-world problems with challenging characteristics like non-differentiability, local optima, and multiple objectives. With AutoEC, we will ensure that even non-expert practitioners can fully exploit the power of EC. Key to our novel AI approach is the repeated application of EAs to problem instances from the same class, as often happens in practice. We will devise learning mechanisms that, across repetitions, can discover relations between problem instances and how to efficiently solve them, so that EAs automatically become ever-more adept at solving these problems. Our main goals are to 1) lay the algorithmic foundations behind AutoEC, 2) build and make available a software toolbox, 3) validate AutoEC in practice. We target real-world applications of high societal and industrial relevance that have the potential for valorization within the runtime of the project. Specifically, we aim at the medical, maritime, automotive, and robotics domain, together with potential societal/industrial partners, including LUMC, Elekta, Philips, C-Job, Qualcomm, BMW, DAF Trucks.

Keywords

Artificial Intelligence, Engineering Design, Evolutionary Computation, Smart Industry

Other organisations

Hogeschool van Amsterdam, Leiden University (LEI), Utrecht University (UU), Vrije Universiteit Amsterdam (VU)

Submitter

Organisation Leiden University (LEI)
Name Prof. Dr. T.H.W. (Thomas) B├Ąck
E-mail t.h.w.baeck@liacs.leidenuniv.nl
Website https://www.universiteitleiden.nl/en/staffmembers/thomas-back