Simultaneous Network Identification and Spatiotemporal Task Planning for Multi-agent Systems

Simultaneous network identification and control has been receiving increasing interest due to its applications to multi-robot systems, intelligent transport, as well as social and biological networks.

Most approaches consider the network identification as a separate problem from the control part, which renders the case of real-time network topology changes problematic.

At the same time, a current trend in multi-agent systems is to target complex goals by employing tools such as temporal logics from formal verification, in order to specify more general specifications that induce a sequence of control actions rather than a stand-alone traditional control objective.

Most existing solutions do not tackle quantitative transient constraints, nor consider the computational cost that may be endured by controlling the whole agent group. At the same time,  the potential uncertainty and lack of knowledge of the network structure is not accounted for.

We consider here an approach that simultaneously handles the network uncertainty and temporal logic task planning problems.

The network identification module runs in parallel with the distributed control one, the latter being adjusted accordingly each time a new estimate is given by the online network identification procedure.

The distributed approach to the problem will guarantee a significant improvement of several attributes, namely scalability, resource usability, as well as robustness wrt link failures and network uncertainty.