Abstract
The Internet-of-Things (IoT) comprises various kinds of devices able to sense, and potentially even interact with, their environment through a variety of capabilities. Small robots are only one group of such devices able to interact with and move through the environment. With the rising numbers of devices expected in the near future, deployed in our highly dynamic environment, a central controller will become infeasible. In the absence of a central controller, individual robots have to operate within their environment and interact with other robots autonomously. To deal with rapidly unfolding situations, robots have to adapt themselves and their behaviour during runtime to changed conditions. The interaction with other devices can lead to emergent, self-organised behaviour of the different devices allowing the collective to achieve common goals more efficiently. In this talk, we will look into the multi-task k-assignment problem where tasks arise dynamically at runtime. The robots only know about the tasks if they are within their sensing range and require multiple collaborating robots to be completed. This leads to complementary goals for the individual robots which need to decide whether to remain in their location and try to detect new tasks or to follow and provision a specific task. I present a novel, goal-aware approach to affiliate each device with a team, pursuing a specific task, at runtime. In this novel approach, devices only operate on aggregated information from the collective which constantly changes during runtime. Finally, I will also present an approach to select a team affiliation during runtime using machine learning techniques.