Hierarchical Distributed Formation Tracking of a Moving Target with Consensus-based Position Assignments and MPC
Motivation
Consider a group of autonomous agents (e.g., mobile robots or UAVs) tasked with tracking an unknown moving target while maintaining a prescribed formation around this target. The target motion is not known a priori and may change rapidly over time. Additionally, communication between agents is restricted, and control and collision avoidance constraints must be taken into account. A naive approach, which assigns roles based purely on instantaneous proximity to formation slots, can result in frequent switching and Zeno-like behavior, particularly when lower-level controllers are model predictive controllers (MPCs), which are sensitive to reference jumps. In contrast, consensus algorithms enable distributed decision-making by allowing agents to reach agreement on role assignments through local communication while ensuring consistency across the network. This motivates the development of a hierarchical, distributed control framework in which a discrete consensus-based role assignment avoids rapid switching and is decoupled from continuous trajectory tracking via MPC.
Task description
To avoid a mixed-integer MPC problem, the thesis focuses on developing an upper-layer role assignment procedure combined with a lower-level MPC-based tracking approach. Hereby, the upper-layer decision-making process should be based on the Min-Consensus algorithm, where possibilities such as additional penalization terms to avoid rapid switching behavior shall be explored. Furthermore, the lower-level local MPC must be designed to ensure appropriate tracking performance. The hierarchical control approach shall be evaluated with simulations where the control quality is to be evaluated with different performance metrics such as formation error, number of switches, communication reliability, and control effort.
Requirements
Basics of MATLAB, Experience in optimization-based control, and/or distributed algorithms (e.g., from the lecture Numerical Optimization and Model Predictive Control)
Contact
Maximilian Pierer von Esch, M.Sc.
Chair of Automatic Control
maximilian.v.pierer@fau.de
Katharina Stich, M.Sc.
Institute for Smart Electronics and Systems
katharina.stich@fau.de