Last month, some members of the PICK-PLACE consortium attended the European Robotics Forum (ERF) 2019, which took place in Bucharest (Romania), on March 20-22 2019. The Forum has been an excellent occasion to learn about the latest innovations in the field, and has provided interesting feedback for our project. Ander Iriondo, researcher at IK4-TEKNIKER, answers the interview.
As IK4-TEKNIKER, why did you decide to attend ERF 2019?
ERF is a reference forum of robotics in Europe, in which experts from different research and industrial centres meet once a year to share their current results and advancements in robotics field.
JW Marriott Bucharest Grand Hotel, where ERF 2019 was hosted.
What did you present in the Forum? Could you explain the feedback you received?
We presented some of the achievements in the context of several projects we are involved in, including PICK-PLACE project. Regarding PICK-PLACE, we took the opportunity to show some deep learning techniques we are applying for (1) infering the most suitable grasping points in multireference parts scenarios, and (2) learning a mobile manipulator base controller for optimal positioning in pick and place operations.
In what ways did you present PICK-PLACE project? Did you receive any feedback?
The work carried out in the the PICK-PLACE project was presented in two workshops: “World perception for autonomous applications in agile production” and “AI, Planning and Robotics: Results and future challenges ”. The assistance to both sessions was of nearly 100 attendants. After the session, several attendants showed their interest in some technical aspects of the solutions.
Ander Iriondo giving a presentation at one of the workshops in ERF 2019.
Which relevant innovations or trends related to PICK-PLACE did you see in ERF 2019?
On the sessions we attended, we had the opportunity to hear several presentations related to learning pick and place operations. One example aimed at learning grasping behaviours using deep reinforcement learning techniques. The OCADO English supermarket also presented an automated warehouse system able to deal with bins full of multireference products.
From the perspective of PICK-PLACE, which ideas do you bring back from ERF 2019?
We have seen that the main trend is the aplication of deep and deep reinforcement learning techniques for approaching the pick and place operations. This is due to the computational advancement that has happened in the last years especially with the improvement of the GPUs. We have seen that our approaches in PICK-PLACE project are aligned with the work of other experts.