Ensemble modeling and control for congestion management in automated warehouses

In this work, we present and evaluate an ensemble approach towards the modeling and mitigation of congestion in multi-agent robotic systems operating in a dynamic environment. In particular, we consider the problem of congestion management between robots and robots and human agents in an automated factory or warehouse. The main contribution of this work is the development of appropriate ensemble models that describe the collective dynamics of the system to identify regions of high congestion. These models are used to design appropriate agent-level feedback control policies to improve the performance of the robotic team. To reduce the frequency of robot-robot and robot-human interactions, the feedback control policies implemented by the individual robots are designed to control the variances in the robot population distribution in the workspace. We present simulation results that validate the feasibility of our methodology.