Collective Construction Using Lego Robots

Crystal Schuil, Matthew Valente, Justin Werfel, Radhika NagpalAAAI 2006

Abstract

This study investigates the feasibility of autonomous collective construction using a swarm of simple, identically-programmed robots. By leveraging "extended stigmergy"—a process where robots coordinate indirectly by modifying their environment—the research demonstrates that complex 2D structures can be built without central controllers or global positioning systems. Using low-cost LEGO Mindstorms hardware and specialized magnetic foam blocks, the authors show that robust collective behavior can emerge from minimalist sensing and local interaction rules.

Introduction

Traditional automated construction typically relies on high-precision robotics, expensive sensor arrays, and centralized "master" computers. While effective in controlled environments, these systems are inherently fragile; a single failure in the central controller or an unexpected environmental shift can halt the entire process.

In contrast, social insects like ants and termites build massive, intricate nests using only local information and simple behavioral rules. This paper translates these biological principles into a robotic system capable of building user-defined structures using "expendable" hardware. The goal is to create a resilient alternative to centralized planning that does not require perfect movement or constant wireless communication.

Hardware and Mechanical Design

The researchers developed a prototype using LEGO Mindstorms (RCX). To overcome the limited input/output ports of a single RCX unit, each robot utilizes two units communicating via infrared.

The LEGO Mindstorms robot prototype featuring a pneumatic arm and magnetic foam block.

The robots employ a pneumatic arm, powered by an onboard pump and air tank, to manipulate building blocks. A key innovation is the "intelligence in the materials": the building blocks are constructed from foam with embedded magnets. This design allows the blocks to self-align during placement, compensating for the inherent mechanical "slop" and inaccuracy of the LEGO motors. Navigation is achieved via two simple light sensors that allow the robots to distinguish between white building blocks and the dark floor, enabling "perimeter following" behavior.

Methodology: Extended Stigmergy

The core of the coordination is stigmergy, where robots "communicate" by changing the physical state of the structure. The authors utilize "extended stigmergy," which involves adding information to the environment. In advanced iterations, blocks are equipped with RFID tags to store state information, simplifying the logic required for robots to determine the next placement site.

Every robot follows the same set of simple rules:

  1. Locate the block supply station.
  2. Find the existing structure.
  3. Navigate the perimeter of the structure.
  4. Identify a valid attachment point (e.g., a corner) and place the block.

Experimental Results

The system successfully demonstrated autonomous construction of 2D structures. The robots proved capable of four critical behaviors: locating the station, finding the structure, navigating the perimeter, and recognizing corners.

Sequence of autonomous construction showing robots navigating and placing blocks to form a structure.

The results highlight the robustness of decentralized systems. Because the robots are identically programmed and act independently, the swarm is scale-invariant; robots can be added or removed mid-build without disrupting the overall logic. The demonstration proves that complex collective behavior does not require advanced computer vision or GPS, but can emerge from simple "dark vs. light" sensing.

Conclusion and Future Work

While the LEGO-based system serves as a successful proof-of-concept for low-cost hardware, it is currently limited to 2D structures. Future research aims to extend these algorithms to 3D construction, mirroring the complex mounds built by termites. Additionally, the authors intend to explore environmental interaction, where the structure's final shape is partially determined by the surrounding terrain, and to test the limits of swarm density with larger numbers of robots.

Summary updated

BibTeX
@inproceedings{schuil2006collective,
  author = {Crystal Schuil and Matthew Valente and Justin Werfel and Radhika Nagpal},
  title = {{Collective Construction Using Lego Robots}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2006},
}
Figure