Distributed Nonlinear Model Predictive Control in Real-time

Abstract: Distributed Model Predictive Control (MPC) aims to transfer the success of centralized MPC to cyber-physical systems. Cooperative approaches with multiple communication rounds between coupled subsystems promise high performance by solving a centralized optimal control problem online. However, real-time requirements only allow for few optimizer iterations and the resulting suboptimal control inputs reduce performance and complicate the stability analysis. Similar challenges have been successfully addressed in centralized Nonlinear MPC (NMPC) by Real-Time Iteration (RTI) schemes. We present an RTI approach for distributed NMPC via tailored decentralized optimization algorithms and obtain novel stability guarantees for coupled nonlinear systems. Experimental results with robot formations demonstrate the efficacy and real-time feasibility of the proposed approach.

Bio: Gösta Stomberg received the M.Sc. degree in mechanical and computational engineering from the Technical University of Darmstadt, Germany, in 2019. Since 2020, he has been a Ph.D. student at the Institute for Energy Systems, Energy Efficiency and Energy Economics at TU Dortmund University, Germany. His research interests include distributed optimization and model predictive control.