Bespoke Digital Fabrication
Design and assembly of lightweight metal structures Gramazio Kohler Research, ETH Zurich 2014-2018
Additive Fabrication Carbon Fiber Wiring
Additive Robotic Fabrication of Complex Timber Structures, Zurich, 2012-2017. ICD/ITKE, Pavilion, Stuttgart, 2014.
Robot Experimentation
The spread of multi-functional industrial robots has had exponential acceleration in the last decade that it has become a standard tool in many industries, where automation, efficiency and accuracy are the heart of the production process. Since the ‘80s, with the development of information and computational technologies, the machines begin to be controlled through digital tools defining the beginning of a new paradigm in which the virtual world and the physical world come together and influence each other, drawing the foundations for the emergence of new processes and production strategies. The development of these technologies does not occur evenly in all disciplines and production companies but condenses on specific sectors, such as automotive and aerospace, going to redefine the quality standards of their respective sectors. The birth of these machines is the result of a need of greater control over the production process and greater automation, elements not necessarily connected with the world of construction and architecture. The diffusion of industrial robots in these fields occurs in manners much slower and gradual as for digitization, because of an industry in which the development timings and implementation medium assume different sizes and durations, a much more fragmented and complex industry.
 Very significant reason is also a limited and well-established design methodology and dominant constructive that has not evolved with the same speed of other disciplines; in defence of this, it is the fact that the architectural project encloses a more difficult to calculate or in some cases of not calculable variable number that can not be foreseen in advance and then inserted in a fully automated process, but they provide, at least for now, the presence of the human factor as an element of conjunction with the real world, able to use intuition and take charge of the major design decisions. Despite a more gradual evolution of the last decade we have seen a substantial increase in the use of robot technology both within the construction process and within the design phase. The flexibility of robots such as industrial arms provides a wide spectrum of potential uses, not limiting them to unique automation tools predetermined and finite processes, but added elements able to expand the possibilities of the designer.  Fabio Gramazio and Matthias Kohler, The Robotic Touch: How Robots Change Architecture, Research ETH Zurich 2005-2013.
RobotIC Experiment 01
Location: ODICO Formworks, Odense (DK) Robot: ABB IRB 6620 (suspended) Objectives: Main objective of the workshop was to test the behaviour of polymorph plastic through the use of a custom made extruder connected to an ABB robot. Testing was aimed to achieve both linear and spatial geometrical configurations and patterns. The workflow has been set up in this way: robot path definition via CAD software, translation into G-Code via in house software PyRabbit, manual or automated execution of the code. Preliminary Phase: Manual control of the robot with linear and spatial extrusions generating catenary geometries as result of the material behaviour and extrusion strategy. Automated Phase: Automation of the extrusion process and calibration of printing speed according to the material behaviour.
RobotIC Experiment 01
The robot setup at ODICO in Odense, Danmark gave us great opportunity to experiment with material behaviour and execute first robot movement tests in jogging mode but also autonomous via G-Code. We rapidly realized that we would not need to move the endeffector in a catenary way to creat arches, but rather move it in a two dimensional manner in the XY-Plane. In that way we can use the materials flowing behaviour through gravity, let the extruded PCL drip and let time form an catenary by itself, as an form-finding, not an form describing process. Variation in time, frequency of stops and movement across supports gave us insight into material behaviour. Throughout the time at ODICO we increased supports, starting with a single linear support, through a second parallel and finally three linear support beams forming a triangle, adding degree of complexity to the extruded catenary dripping process.
Coded Robot Path Catenary Material Formation
RobotIC Experiment 02
Location: University of Innsbruck Exparch-Hochbau, (AU). Robot: ABB Agilus Objectives: Main aim of the workshop was to print a designed piece that was geometrically, structurally or systematically related with the studio project, by using the facilities of REX LAB (Robotic Experimentation Laboratory). One of the 3 ABB Robots was used to print concrete with a custom made end effector developed by Hochbau department. The setup consisted in a pump mixing cement and water, transporting the aggregate to the enedeffector and a smaller pump to inject a chemical eccelerator. The calibration of robot speed, endeffector deposition speed, material consistency resulted in a delicate operation. First Session: Introduction and Tests Introduction to the software that is used to control the robots and setting up the workflow, from C-Code to G-Code. Second Session: Printing Robot path check in Robot Studio, sequence check with the robot and final deposition of material.
Two different approaches have been explored to generate the geometry and robot path file; the output files have been tested with a custom robot simulation developed in C++. These two methods were both tested and their output robot path files were used to generate the input G-Code for the robot control. 1 Model and path generated through C++ code. 2 Model definition in Maya and path file preparation through python scripts. The model has been firstly sliced with a 0.75 layer heigth and then the coordinates exported in .txt file format. 3 The output file of the two approaches has been input in the robot simulation to generate G-Code. 4 Robot Studio has been used to read the G-Code, test robot motion, clash detection and execution.
C++ _Model _Points generation Maya _Model _Slicing (Python Script) _Points Generation (Python Script)
2 Slicing Geometry _Maya python code: _Select 3 vertices of a reference plane and execute _Select the geometry and execute _Layer Height: 0.75 cm
Concrete printing end-effector Resolution / Texture
Endeffector Design
Endeffector 1.0 Next stage of the research was to develop a custom tool for the robot to operate in an automated way. The experience gained with the PCL extruder throughout the workshop in ODICO led us to design an early version of PLA extruder to start experimenting with the variables involved in the robotic fabrication. The end effector was developed for a Nachi MZ07 robotic arm and allowed to experiment with plastic extrusion within the studio environment. Extrusion speed, waiting time and robot speed are variables that have been experienced and tested.
Endeffector 2.0 Main weakness of the former experiments was the connectivity between members. The imperfect nature of the plastic rods influenced consistently the correct positioning in space causing main deviations from the digital model. The assembly was also difficult because the reduction of material due to the melting in the ends of each rod. The propagation of deviations throughout the entire model caused difficulties in scaling up the fabrication to more then few members. The use of a second material, (PCL) has been considered as a gluing system between members without having to locally melt the rods themselves.  This required the development of a new end effector which embedded a cutting device and a second extruder for the deposition of the gluing material (PCL) with the aim of a full automation of the entire fabrication process. Further decision has been to move all the electronics offboard in a dedicated control panel which allowed more manageability and more direct control of the extrusion speed controller.
Endeffector 3.0 The experiments conducted with the former end effector setup underlined a the necessity to simplify the entire sequence. The end effector implied the execution of complicated operation which made each step of the entire sequence not reliable and not solid enough for an automated procedure. Further step in the development was to re-evaluate the entire sequence fragmenting the single operations of the end effector in off board and on board steps. The extrusion and cutting phase was moved off board in a dedicated station with the aim of deploying discrete elements with custom lengths in a more consistent way. The third generation of end effector was a gripper tool utilised to collect and locate in space the discrete elements previously prepared by the cutting station. The assembly of multiple elements happened with the help of manual melting with a soldering tool. A second robot was considered to work in a collaborative environment with the assembly one. The experiments carried on clarified the feasibility of the process with more precision and consistency of the formers. The proxy material utilised (PLA), guaranteed an agile experimentation with a Nachi MZ07 robotic arm, due to the lightweight material properties. The robotic arm precision and capability to keep location in space for unlimited time has been utilised within the process to negotiate with the phase changing properties of the material in use. The fabrication sequence explored, showed the possibility to scale up the process to bigger machines and move forward to a stiffer material: Mild steel. 3mm rods were utilised for further manual experiments using spot welding as technique to assemble the discrete elements, resulting in more stiff and rigid models.
The fabrication sequence is divided into four main moments: Bespoke cut of discrete elements: a cutting station has been designed in order to deploy segment with custom lengths according to the structural need. The segments are cut and kept in place within the station to allow the robotic arm to record the position and automate the gripping process throughout the sequence. A proximity sensor has been installed to actuate the cutting process according to the position of the robotic arm and give indipendency from external actors. Gripping: a custom end-effector has been developed with a linear gripping function in order to pick the discrete elements. The actuation of the tool happens through the digital output of the robot; the connection end-effector / digital output allows to have the gripping process indipendent from an external controller or actuator, increasing the consistency and reliability of the whole process. Spatial positioning: the segments are gripped, positioned and oriented in place according to the network configuration. Assembly: the segments are assembled togheter through the collaboration of a second robotic arm with a soldering tool. The second robot is actuated in sequence after the positioning is completed. Main objective is to keep the sequence as consistent and solid as possible as well as simple, with the aim of a full automated assembly process. The sequence developed is mostly scale indipendent and could be migrated to bigger robotic arms for a 1:1 application. Spot welding has been thought as process of assembly with the use of metal rods instead of plastic. The use of heavier rods requires a more reliable endeffector with higher torque force motor, more solid materials and a double gripping unit to avoid the rod instability problem during the robotic movement. A fourth generation endeffector was designed to counter these issues and propose the further step of the development.
 Robotic arm setup with off board controller.
 Cutting station to deploy segments with custom length.
 Linear gripper to position and orient the discrete elements.