MAchIne learning for Design of industrial Robots MAID-Robot
|Coordinator||Linköpings universitet - Institutionen för ekonomisk och industriell utveckling|
|Funding from Vinnova||SEK 500 000|
|Project duration||November 2018 - November 2019|
|Venture||Groundbreaking ideas in industrial development|
|Call||Banbrytande idéer inom industriell utveckling 2018|
Purpose and goal
The goal of the MAID-robot project is to drastically reduce the development time for industrial robots through the use of machine learning based on virtual simulation models and large amounts of customer data. The project has shown that it is possible to automatically configure individual robots using statistical methods to run an optimization framework based on how customers actually use the robot in their production today.
Expected results and effects
The MAID-robot project has created a framework that enables the automatic assembly of a virtual simulation model, a digital twin, of the industrial robot´s drivetrain and then allow an optimization algorithm to vary its parameters so that it can best solve the tasks that customers use the robot for today. Thus, a product can be developed that fits the customer perfectly in a shorter time than today.
Planned approach and implementation
** Denna text är maskinöversatt ** The project´s first work package consisted of conducting a state-of-the-art analysis, and based on this, a digital framework was developed to put together a virtual simulation model of the drivetrain of an industrial robot. The next step was to develop a search algorithm where machine learning and optimization are linked so that the optimization algorithm searches for solutions that best meet the requirements that the machine learning algorithm has identified. Finally, the framework has been implemented and tested in ABB´s software.