Factory Floor: A Robotically Reconfigurable Construction Platform Kevin C. Galloway, Rekha Jois and Mark Yim Abstract— Passive robotically-reconfigurable truss structures offer considerable utility as they can quickly adjust to changing functional requirements and resources at a level of sophistica- tion that no human builder could match. Furthermore, robot built structures can be constructed in environments such as surface of Mars or in micro-gravity, which would otherwise be too time consuming or dangerous for humans. In this paper we discuss some of the mechanical design challenges of developing a passive robotically-reconfigurable truss system, and present the concept of the factory floor, which can construct truss-like structures without climbing on them. In the proposed system, each level is constructed on a ground plane using a truss and node configuration and is elevated to make room for the next level. This process is repeated to create 3D truss structures or reversed to decompose the structure for the next task. I. I NTRODUCTION AND MOTIVATION Much can be learned from abstraction of behaviors or properties found in nature and their implementation into robotic systems at the mechanical and control level. The robustness of natural systems to compose and decompose elements offers benchmarks for the robotic assembly and disassembly of synthetic structures. Application areas for robotically assembled structures have focused on large space structures though deep-sea mining and martian or lunar structures are other possible applications [16] [14] [18]. If these structures could be taken apart, they could be reconfigured to adapt to needs as situations change. Towards this longer term goal, this paper focuses on the electro-mechanical platform for which algorithms can then be developed to autonomously (de)construct structures as needed. One example algorithm develops local reactive behaviors that result in the ability to ”robustify” the assembly of this distributed platform [12]. In view of these motivations the design aims to allow distributed computation, manipula- tion, sensing towards a larger coupled structure with minimal interaction constraints and cost. Note that all processes here are reversible and so assembly processes equally apply to disassembly. II. RELATED WORK While there have been many robot hardware platforms for a variety of robotics research[11], there have not been many platforms for autonomously assembled structures. In [17], a modular structure is proposed in which an assembler robot puts together structures made of cubes. The assembler robots climbed on top of the structure building what looks like a brick wall. While this paper proposes a design that This work was supported by NSF EFRI The authors are with the department of Mechanical Engineering and Applied Mechanics, University of Pennsylvania, Philadelphia, PA, USA kcg@seas.upenn.edu; yim@grasp.upenn.edu Fig. 1. A simple illustration of truss-like construction where items in red are trusses, items in blue are nodes, and items in gray represent the robotic assemblers. can form 3D structures and has passive cube modules the implementation had active cubes and could only implement 2D structures. Other work has focused on the control of robots that assemble trusses[3]. Developing a robotic system capable of assembling truss- like structures presents a very coupled mechanical design challenge. The robot, or builder, can not be fully realized without considering the geometry of the truss and the means of connecting trusses together. Likewise the truss design can not be finalized without fully understanding the limitations of the builder. Prior work in robotic construction of static structures has predominately focused on enabling a robot to navigate a structure and manipulate (i.e. reorient and place) the structural elements. Shady 3D [18] was designed to be an active mobile module capable of manipulating passive structural modules within a 3D truss structure; however, this robot lacked the ability to physically attach the passive structural modules to the structure. Hjelle and Lipson [9] developed a “hinge” robot capable of traversing simple cubic truss structures as well as removing and docking trusses. A passive truss configuration was presented whereby a rod with threaded ends could be inserted into one of 18 threaded sockets of a node, which acts as a connection point for multiple trusses. Two rods are required to the span the distance between two nodes, which is accomplished by partially unthreading the two rods from the center to thread into adjacent nodes. While truss manipulation and attachment were demonstrated more work is still needed to enable to robot to manipulate and attach the nodes. Another method for achieving a desired shape is through robot self-disassembly by starting with a collection of robots in an amorphous 2010 IEEE International Conference on Robotics and Automation Anchorage Convention District May 3-8, 2010, Anchorage, Alaska, USA 978-1-4244-5040-4/10/$26.00 ©2010 IEEE 2467