Search Space Illumination of Robot Swarm Parameters for Trustworthy Interaction

For swarms of robots to become commonly adopted, operators will need to be able to adjust swarm behaviour in an intuitive and simple manner. To achieve this, we produce a feature space of swarm characteristic sets using Map-Elites. These sets provide users with the ability to explore specific combinations of characteristics known to achieve good results for the specified requirements. This allows operators to pass preferential parameter sets to swarm agents, without requiring a deep understanding of how parameters effect swarm behaviour. The resultant system gives the user reassurance that the swarm will operate to their needs while maintaining the benefits of distributed swarm technology. We demonstrate how this system might be used with a swarm logistics case study where an operator requires two swarm behaviour modes with differing characteristic needs. In this example we illustrate how an operator might express preference, receive recommendations, and select their final swarm characteristics.