The implementation of the DEEPFIELD concept will be done by means of two complementary set of coordination and support activities with the aim of contributing to the two ultimate goals of this twinning action: 1) raise the scientific knowledge of INESC TEC researchers in key areas of deep-learning in field robotics; 2) improve the INESC TEC researcher profile in deep-learning applied to field applications.
Coordination activities will materialize what has been described in the five major actions from the DEEPFIELD concept, such as:
Supporting activities will be side activities that will support and enhance the major pillars of the project.
Five main actions support this overall concept:
A training strategy based on sessions touching deep-learning robotics aspects followed by sessions focused on specific fields of expertise, allowing researchers to improve their knowledge in a specific field of expertise while taking into account the context and interrelations with other fields of expertise and preparing them for the research and implementation challenges in the field with harsh environment. From this pillar, 4 thematic workshops will be promoted.
Short period of time (2-3 months) where researchers are invited to stay in the partnering institution and work with them in order to learn, by experience, the acquired theoretically in the thematic workshop. From this pillar, 4 short-term scientific missions (STSM) will be promoted. This activity will be mentored by an assigned tutor selected by the counseling board.
A training strategy targeting hands-on training will be pursued, enabling direct application of the knowledge acquired in theoretical sessions and in the short-term scientific missions and improve their understanding of the actual requirements of the stakeholders, allowing them to design solutions with higher economic potential. From this pillar, 4 hands-on summer/winter schools will be promoted.
A new collaboration network involving top EU research institutions in the deep-learning area, including the key fields of expertise required to address the challenges identified above, articulated with existing networks in specific fields of expertise, such as EIT, euRobotics, ETP NetWorld2020 and Raw Materials. From this pillar, networking meetings will be promoted.
Interchange and interaction among different fields of expertise is fundamental to address the deep-learning robotics challenges (such as, grasping, semantic mapping, multi-robot cooperation, etc.), due to the multidisciplinary nature of the application area. The interchange and interaction between researchers from different institutions and different fields of expertise will therefore be targeted. From this pillar, conferences and scientific events will be attended and promoted.