Model-based metacontrol for autonomous robots

In 2016, Team Delft won the Amazon Robotics Challenge, a competition on autonomous manipulation. These competitions reveal the challenge in integrating complex robotic system architectures that need to perform complete tasks in real environments. Methods and tools are needed to build robotic systems that are reliable and trustworthy. The knowledge of the system’s design is what enable engineers to operate, re-configure and adapt such complex partially autonomous systems in mission-critical scenarios (e.g. space).

Model-based Systems Engineering enables to capture the engineering design of the system in a single model.

Our metacontrol framework leverages that design knowledge at runtime for self-adaptation, by using a metamodeling approach to obtain a runtime model in the form of an ontology, upon which automatic reasoning is applied for diagnostics and reconfiguration of the system’s architecture during mission operation.

Within the ROSIN European project, we are developing an implementation of this framework for ROS systems. This is only one of the ROSIN actions, which aims to foster ROS-Industrial as a European Digital Platform for Robotics ready to meet industry demands. These include: funding to 3rd parties to support ROS open source development, promoting Quality Assurance in the ROS community, and Education activities, such as industrial trainings of the Hello (real) World with ROS MOOC.