Don't let player get stuck!: Increasing the autonomy of agents

Abstract

There is still a lack of methods to build multi-robot systems that might be easily modified or expanded on new functional blocks. Moreover, one of the most popular paradigms in robotics, the behavior-based control, despite of its ability to provide architectural transparency and intuitive means of describing the behaviors of robots, does not allow us to predict final quality and performance of the system. To increase safety of robots and the sustainment of their operations, for example, we can consider all the potential of sensors. In this paper we present current state of our multi-agent architecture, exhibiting their reactive properties. We present a simple trap recovery behavior, which can be executed by robots in a case of abnormal situations, e.g. hardware malfunction. On the other hand, we specify a multi-robot construction domain, which will serve as a scalable testbed for our control strategy.