Navigation-based optimization of stochastic strategies for allocating a robot swarm among multiple sites

We present a decentralized, communication-less approach to the dynamic allocation of a swarm of homogeneous robots to a target distribution among multiple sites. Building on our work, we optimize stochastic control policies for the robots that cause the population to quickly redistribute among the sites while adhering to a limit on inter-site traffic at equilibrium. We propose a way to account for delays due to navigation between sites in our controller synthesis procedure. Control policies that are designed with and without the use of delay statistics are compared for a simulation in which 240 robots distribute themselves among four buildings.