INTRODUCTION

The introduction of digital tools and associative modeling into design in the last few decades allowed architects and designers to easily create complex structures and geometries [1]. Yet, when coming to realize these complex structures in the ‘real world’, we notice a big gap between these digital processes and their construction processes, as the majority of the building industry still uses standard tools. The ‘easiness’ that characterizes the digital design process fades away and the process becomes tedious and ad-hoc, often demanding some painful compromises in the final outcome.

Inside that realm we wish to tackle a specific issue of on-site assembly of digitally generated complex structures (in our case double-curved self supported structures). In order to realize these structures in the ‘real world’ we often use strategies of discretization and panelization of elements. Therefore, the assembly process is one of the most crucial and complex moments in the construction workflow.

In this paper we propose to rethink the design-construction workflow by providing geometric information of the design model to an on-site robotic arm that performs the assembly process. The robotic arm is then part of the design process from its early stages, which in turn account for a more continuous and smooth design-construction process.

In recent years we have seen multiple projects that tackle the issue of robotic assembly in architecture. The majority of them involve vertically brick stacking, where a minor change in the orientation of each brick allows the creation of seemingly double curved or complex structures. This strategy, though impressive and novel, enforce a great amount of constrains on the geometry and on the final outcome [2]. We propose an alternative method for assembly that enables the construction of double curved self-supporting structures.

In many industries and for many years now, industrial robots have been crucial components in complex assembly lines  (e.g the car industry). As the ability of the architect/designer to easily generate complex structures grows, the demand for an analogous assembly technology in the building industry is evident. We believe it is only natural to match the robotic abilities with on-site assembly. The automation of the assembly process will decrease the gap between the design and the construction processes, and between the initial architectural model and the final outcome.

METHODOLOGICAL PROCEDURES

Design from robot constraints

The first step in incorporating the robot in the design process is to understand its abilities and constraints. The robot manufacturer will normally provide this information, but at the moment this is not compiled in one accessible, and interchangeable open format. Therefore, it is necessary to make some iterative tests to evaluate the interaction between the desired shape and the robot capacities in a robot programming environment. In this project, we used HAL plugin for Grasshopper to determine the maximum robot envelope space. Also, we have used a trial and error evaluation to determine the tool’s lifting capacities such as maximum dimensions, maximum weight and appropriate surface finishing of the pieces to be assembled. Once the overall scale of the structure has been fixed relative to the maximum reachable point of the robot’s envelope space, we are able to subdivide the shape in smaller parts according to the tool’s maximum weight capacity. This final step, the discretization of the initial shape, can be achieved by any diagrid panelization algorithm. In our project, we used a panelization method based on Pier Luigi Nervi’s diagrid structures.

Structural scheme and sequence of assembly

Merging two typologies - the tower and the bridge - the subdivided mushroom structure works in pure compression, tightened with a tension ring at the top. In order to connect the resulting panels and to assure structural efficiency, we created a male-female cantilevered joint. The result was a self-supporting structure that relies on continuous compression of each element. In order to enable the smoothly sliding of each piece into another, we made several tests and evaluated the accuracy range of the robot. In accordance with this, we defined the appropriate tolerance of the cantilevered joint. The structure was assembled ring by ring, which assured it will be self-supported at each of the assembly phases, and omitted the use scaffolding.

Method of assembly

The next step is to match the existing geometry with the information that the robot can read. First, we need to find a suitable toolpath (a description of the robot’s movements in space) for the assembly of each piece. We extracted and extended an isocurve from the piece and used it as an input for the programming sequence so that the robot moves in the same uv plane of the given 3D shape. This simple matching between the geometric information and the robotic arm toolpath assures a smooth fit and also guarantees no overlapping between neighboring pieces. Therefore, the geometric information is being used as an input in the programming sequence so that the robot motion is finally described through sequential linear movements.

On-site assembly

For the actual assembly of the pieces, we attached a KUKA KR10 to a small trolley to allow the movement and positioning of the robot on-site. The location of the robot was fixed during the whole process of assembly, so in order to reach every point of the structure it was necessary to develop a rotation method of the target structure. We developed a simple and accurate rotation system by attaching a turning plane with a fixed rotation of 60 degrees to the bottom of the column. Finally, the code of assembly was set to perform the assembly of 1/6^th of the structure at once, and then it was repeated 5 times to complete the full circumference of the structure.

RESULTS

The result of this project is a 1.40m/4.6ft tall, 3m/9.8ft wide self-supporting structure assembled on-site by a KUKA KR10 with a customized vacuum lifter as an end effector. Additionally, we developed a code that translates geometrical information of any diagrid structure into machine movements or toolpaths. Finally, we develop a female/male joint to assemble any thin double curved cantilevered structure that can be easily fabricated with any 3-axis CNC machine.

DISCUSSION

In this project we:

- Developed a digitally generated geometry that allows an on-site robotic assembly of double curved, self-supporting structures without using scaffolding,

- Implemented a method for translating the structure’s geometric information to inputs in the robot code,

- Hopefully demonstrated that the assembly of complex structures can be as easy as generating these structures in a computer.

With our contributions we aim to question the scale and processes involved in ongoing research of on-site construction assemblies [3]. While current studies focus on the development of complex geometries and artifacts, usually in the form of pavilions or medium scale installations, in this project we focus on the development of a customized but standard building component, a mushroom column that can be implemented in any type of buildings.