Abstract
Traditional robots have fixed shapes, limiting their ability to adapt to complex or dynamic environments. Modular robotic systems are composed of identical robot modules that can connect and disconnect to form different structures. This allows the robots to form bridges to cross voids and split apart to fit into tight corridors. However, while previous modular robotics research heavily emphasizes reconfiguration planning, comprehensive motion planning remains underexplored. We introduce a novel, end-to-end framework for the planning, state estimation, and control of chain-style modular robot swarms. Our system includes a custom network flow planner that maps polygonal environments into a Reeb graph over chain length: for each integer length, we compute the traversable free-space topology, and the graph captures how corridors open or close as chains grow or shrink. By bounding graph edges with void-spanning minimums and capacity maximums, the routing problem is formulated as a MILP that remains tractable as swarm size increases. The MILP generates a high-level sequence of waypoints and reconfigurations, which are executed using an Unscented Kalman Filter (UKF) for overhead state estimation and ORCA for collision-free local navigation. Our approach integrates topological planning with low-level control, enabling modular chains to autonomously navigate environments featuring voids and tight corridors.
Document Type
Article
Class Name
Electrical and Systems Engineering Undergraduate Research
Date of Submission
4-21-2026
Recommended Citation
Theiler, Sebastian; Nie, Yucheng; and Kevil, Josh, " Planning, Control, and State Estimation for Chain-Style Modular Robots" (2026). Electrical and Systems Engineering Undergraduate and Graduate Research. 54.
https://openscholarship.wustl.edu/eseundergraduate_research/54