Cognitive Control

Work Package 5 - Cognitive Control

The goal of this work package is to develop high-level and monitoring technologies that can be used to improve the performance of the system in different aspects such as: flexibility, safety, etc. Powerful tools must be developed and implemented to solve not only very complex problems such as planning and unloading actions, motion planning and collision avoidance, but also to enable the system to learn performance improvement through learning-from-experience. The methods developed in this WP will provide automatic mission control and supervision as well as user interfaces that assist the human operators efficiently interact with the whole system. The RobLog system needs to make decisions on the basis of the surrounding environmnent:

1. Which object should be unloaded next
2. How to grab it while avoiding collisions with the other objects in the scene?

To begin, the first decision si made by the Hybrid Planner module, which funtions as the core decision-making tool of the RobLog demonstrator. An item is judege as a good candidate to be unloaded next, it its removal from the unloading scene does not impact the static equilibrium of the other items.

The interpretation of the scene is based on geometrical and static equilibrium analyses enabling the symbolic representation of the scene. Such representation graphically highlights the hierachical relation between items in the scene: just like a pyramid, the less important item is the one that does not support any others. This symbolic representation is used for selecting the next object to be unloaded, simply respecting the rule of selecting the "less important item in the hierarchy", specifically the one that does not support any others.

Work Package 5 Deliverables