DeRP algorithm lets robot swarms self-assemble power grids at 125% optimal length

Deploying sustained power in unstructured field environments usually requires pre-planned wired networks or heavy batteries, both logistically painful. A new arXiv paper from Mohammadali Rashidioun, Sangwoo Park, and Petras Swissler introduces DeRP (Dendritic Recursive Pivoting), a decentralized framework that lets robot swarms self-assemble power-delivery networks from a common source to multiple sinks. Robots act as conduits, using only local communication and bearing-based sensing toward sinks, so there's no need for global planning or a map. The algorithm grows branches recursively at locally selected pivot points that approximate Steiner points, efficiently routing to multiple sinks while keeping the network adaptive and scalable.

In quantitative tests, DeRP-formed networks asymptotically reach about 125% of the global minimum length (compared to a theoretical Minimum Spanning Tree), while cutting power losses to 65% of what a Euclidean Steiner tree would produce—a meaningful improvement for field deployments. The team also tracked simulation completion time as sink and robot counts grew, finding sub-linear scaling for up to 100 sinks, meaning the approach stays computationally practical. This positions DeRP as a promising solution for resilient, adaptive power delivery in disaster response, remote operations, or other environments where traditional infrastructure is impossible to pre-establish.