Generalized multi-agent systems can be evaluated and benchmarked in stylized search and task allocation problems (STAP) [1]. In STAP agents search for tasks and recruit other agents to help solve and complete the discovered tasks. Multi-agent coordination algorithms, like swarm intelligence and game theory, are of special interest when assessing multi-agent system performance in STAP. However, these algorithms do not fully generalize all aspects of natural cooperative systems, especially in the case when agents are self-interested. In this thesis the master candidate would investigate how humans would guide and steer agents as avatars in STAP. By controlling for relevant system parameters, like inter-agent communication, agent computing power and unit design cost, novel coordination algorithms could possibly be discovered when analysing the avatar behaviours. The assignment requires the development of a suitable man-machine interface for human steering of agents in STAP, possibly using online recruitment of humans for analysis of massive swarms.
[1] M. Minos-Stensrud, H. Jonas Fossum Moen and J. Dyre Bjerknes, "Information sharing in multi-agent search and task allocation problems," 2021 IEEE Symposium Series on Computational Intelligence (SSCI), 2021, pp. 1-7, doi: 10.1109/SSCI50451.2021.9660121.