Introduction: Minibuilders - How to 3d Print Big Structures With Small Robots
There has always been a close relationship between architecture and technology. However, recently architecture has stagnated and the construction industry has been slow to adopt technologies that are already well established in other fields. Whilst we design digitally we still construct manually.
Robotics offers great potential towards innovation within the construction industry. However, in their current form applied to the architectural field, in particular construction robotics, these systems all share a specific limitation: the objects they produce are linked to and constrained proportionally to the size of the machine. This methodology of production and construction is not scalable. In this sense, to create a house, using current construction robotics, the machine needed must have a work envelope as large as the house itself.
Hence, the project here below elaborated aims to address this particular limitation through the creation of a technology that is both scalable and capable of fabricating structures using tools that are independent of the final product’s shape or size. The Team explored and investigated the potentials of additive manufacturing (3D Printing) applied to the architectural scale.
Substituting one large robot for number of smaller more agile robots, we developed a family of small-scale construction robots, all mobile and capable of constructing objects far larger than the robot itself. Moreover, each of the robots developed was to perform a diverse task, linked to the different phases of construction. Working together as a family towards the implementation of a single structural outcome.
Minibuilders is a family of three robots, each robot linked to sensors and local positioning system. These feed live data into custom software allowing control over the robots movement and deposition of material, an fast setting artificial marble.
Current research focuses on 3D printing for the FDM (Fused Deposition Modeling). Which means that three dimensional objects may be produced by depositing repeated layers of solidifying material until the shape is formed. Material adheres to the previous layer with an adequate bond upon solidification, may be utilized.
The positioning device, the small robots all equipped with the print-head nozzle, move in a predefined path depositing material in layers. Each layer base is defined by the previous layer, and each layer thickness is defined and closely controlled by the height at which the tip of the print-head is positioned in relation to the previous layer. Controller, material supply and power sources are connected externally.
Step 1: Base Robot
The first robot, the Base Robot, lays down the first ten layers of material to create a foundation footprint. Sensors mounted inside the robot control direction, following a predefined path. Traveling in a continuous path allows for a vertical actuator to incrementally adjust the nozzle height for a smooth, continuous, spiraling layer. The advantage of laying material in a continuous spiral is that it allows for constant material flow, without having to
move the nozzle up at intervals of one layer.
The foundation robot size 26*35*37 cm, weights 2.05kg.
Tools and materials:
-Makerbeam with bearings and bolts and nuts
-QTR-8RC reflectance sensor array
-Electronic tape
-Waterjet aluminum gears
-Laser cut acrylic
-Motor, Axle and wheel mount (aluminum or 3d printed)
-9/4/4mm bearings
-4mm metal shaft
The base robot is mobile, with a vertical CNC moving the nozzle up in Z, incrementally while the robot moves along a predefined path. This creates a continuous spiraling toolpath for the nozzle laying the material on top of each layer previously printed.
All the three robots make use of Makerbeams for their frames, giving flexibility within the prototyping process. Makerbeams are reusable and relatively easy to adjust using the T-slot bolt system. The CNC is also made using the makerbeam bearing kit. You can learn from projects listed here.
The connection between the wheels, chassis and motors is simple using a bearing to support the wheel shaft attached with self locking nuts. The stability of the shaft is reliant on the size and spacing of the bearing set. This solution created a more stable platform for printing than a soft suspension system.
For the robot to follow the predetermined path we used a basic QTR-8RC reflectance sensor array. This allows the robot to follow a predefined path. When testing outside we found that the sensor would lose the track of its path under strong direct sunlight. The simple solution we found to this was to shade the sensor from the sun.
The Arduino and Processing file can be found in step 4 software section.
Step 2: Grip Robot
To create the main shell of the final structure, the second robot, the Grip Robot, attaches to the foundation footprint. Its four rollers clamp on to the upper edge of the structure allowing it to move along the previously printed material, depositing more layers. The nozzle moves dynamically allowing for greater accuracy of material deposition. To create a curved surface the material output will be incrementally offset. Heaters, integrated into the robot’s structure increase the local air temperature to influence the curing process. Controlled by custom software the robot follows a predefined path, but can also adjust its path to correct errors
within the printing process. Rotational actuators control height above the previous layer to maintain a consistent layer.
The Grip Robot size 40*27*12 cm, weighs 4.6kg.
Tools and materials:
-Makerbeam
-Dynamixel ax-12 servos *9
-6mm metal shaft
-Springs
-6mm linear motion
-Waterjet aluminum for body and gears
-Plastic gears -18/6/6mm bearing
-3D printed wheel core
-Rubber to cast wheel and rollers
-Heat gun stripped for heat element, sensors and fans.
-3D printed pulley and belt from hobbycar shop
-4mm acrylic for laser cutting
The print-head positioning system comprises of a front-back linear motion system (11),
side linear motion system (12), front-back motion actuation system (13), side motion actuation system (14) and actuators for print-head or multiple print-heads. The connection of print-heads to the material supply is illustrated in (15). The device positioning system consists of feet (5) each attached to a further four or more legs (4) that are attached to the frame. Legs are connected to the frame via a linear motion system (9). Linear motion bearings (10) allow smooth movement relative to the frame. This motion can be actuated either by springs or by linear actuators (8). Feet are attached to legs via rotary joints that allow rotation relative to the legs. This rotation can be controlled and is actuated by the foot actuation system (7). Wheels (3) are mounted on the feet and are moved using the wheel actuation system (6). The surface is coated with a durable, flexible material to reduce vibration of the device during movement and to increase its grip to the structure on which the device is attached.
In the case the printing path is closed without openings, the movement of the robot is continuous and front-back print-head motion may not be used. The device is placed on the footprint at a desired position, once material is extruded through the aperture of the print-head and the device starts moving in the desired direction, print without pausing is to use one continuous spiral path. In this case the wheels are not in their centralized position during the printing process, the device is constantly moving upwards; every complete rotation the device moves up by the height of one layer.
Having the print-head fixed in a fixed position would cause a number of problems, on any curved path the print-head center would deviate from the center of the path, where the center position of the nozzle is marked with a dashed line. In this case each preceding path would deviate accumulatively from the desired shape thus printing a different structure than programmed.
This can be solved by introducing linear side to side motion to the print-head, it can be used to correct the deviation and control the path of the print head. To position the print-head correctly a mechanical sensor calculates the deviation and adjusts accordingly. If the position of the device and path curvature are known - deviation can be pre calculated geometrically without the need for additional sensors.
If the print-head was confined to only following the previous layer, printed structures could only be shaped as extruded versions of their footprints. This would exclude forms such as vaults, and cantilevers. Shifting the print-head provides an opportunity to alter the curvature of the wall during the printing process. When layers shift relative to their previous layers the curvature of the wall changes. The amount of shift should be precalculated, varying dependent on the position of the device. In the case that the position of the robot is incorrect or an error is detected - curve deviation should be taken into account and either added or subtracted from the shift. The position of the robot in relation to previous layers can be abstracted in various ways, for example different types of sensors from Local Positioning Systems to rotation counters attached on the wheels leading to many possibilities.
Step 3: Vacuum Robot
Another major limitation of today’s additive manufacturing techniques is linked to the unidirectional nature of layer orientation, creating an inherent weakness. Additive manufacturing allows for heterogeneous optimized distribution of matter. To take advantage of this, and not succumb to this limitation, we used structural optimization tools to create a second layer of material over the shell. The material is also closely aligned with the direction of stress, finally optimizing both orientation and thickness of the shell structure. The data derived from the structural analysis is then translated into paths for the third and final robot, the Vacuum Robot. Using a vacuum generator this robot attaches to the surface of the previously printed structure. Moving freely over the first shell on its tracks, depositing material on the surface of the shell, enhancing its structural properties. This task can be performed by one robot, or a swarm of robots working in coordination.
Consulting papers below:
http://link.springer.com/article/10.1007%2Fs10846-013-9820-z#page-1
http://jin-shihui.com/minibulders/Schmidt12.pdf
The Vacuum robot size 30*27*12 cm, weighs 2.1kg
Tools and materials:
-Makerbeam
-Dynamixel ax-12 servos *4
-Silver Duct Tape
-Smoothon Ecoflex Silicon 00-10(softest)
-700w handheld cyclone vacuum cleaner. We used the Core.
-Foam for CNC milling
-4mm acrylic for laser cutting
-Motor, Axle and wheel mount (aluminum or 3d printed)
-Rubber Tracks
The frame of the robot is similar to the first robot, an aluminum mounting affixes the motors and the wheel mechanism to the frame. Whilst the vacuum generator is bolted to the frame. The vacuum generator extracts air from below the robot and a flexible suction cup seals the irregular surface. Negative pressure in the space between the suction cup and the surface attaches the wall climbing robot to vertical and horizontal surfaces. In order to give the robot mobility on double curved surfaces, The suction cup should be adjusted to be lower than the tracks ( approximately 3mm). The frame of the robot should be as rigid as possible and the suction cup should be as soft and malleable as possible. There are other solutions to allow the wheels/tracks to always maintain traction, such as using a suspension system but we found this to be the simplest and most reliable.
The force of the vacuum generator must overcome the weight of the robot. Moreover the power/torque of the motors must overcome the overall weight of the robot, plus friction between the suction cup and the surface. Rubber-like materials produce friction with other surfaces, especially when a force is applied. We experimented with many solutions looking for a rubber-like material or coating to reduce the friction. An alternative solution was found by applying one or two layers of silver duct tape onto the suction cup. We found that this significantly reduces the friction and proving more durable than plastic coatings or lubricant. In practice the tape coating would slowly degrade, however it was easily reparable.
For certain curvature(single curvature or double curvature) the size of the wheel, the distance of the wheels are smaller the better, the weight of the robot should be as light as possible, bigger suction cup gives better suction force. So there is a balance between the weight of the robot and the power of the motors, suction cup size, and robot size.
Step 4: Software and Actuators
All the robots are use Dynamixel servos. These smart servos have good torque, also embedded encoders
http://www.robotis.com/xe/dynamixel_en
The first Arduino library was written by Alejando Savage for AX-12 http://savageelectronics.blogspot.co.uk/2011/01/a...
and a new updated version for MX series can be downloaded here: http://savageelectronics.blogspot.co.uk/2011/01/a...
We also used an Arduino Mega plus drivers.
(http://ro-botica.com/es/Producto/Adaptador-Arduino-Bioloid-CDS55xx/)
You can download the processing and arduino files to control the robots with documentation of the servos in our Github With Special thanks to Guillem Camprodon.
Step 5: Material Extrusion
The material used by the robots must exhibit certain properties, strength, bonding strength, low cost and curing time. Curing time is particularly crucial to allow the robot to travel on top of layers previously printed whilst maintaining structural integrity. For the final structure we use a mix of 40% Axon Easymax two component polymer, together with 60% marble powder. Each layer is 6mm high and 16mm wide. The layers are bonded together becoming a uniform shell when cured. Shell curvature can reliably reach up to 90 degrees with a 30cm radius. It is also possible to increase the percentage of marble powder without loosing mechanical properties. However as the viscosity of the material increases the amount of pressure needed to move it through pipes also increases, which in turn makes control over material flow harder.
After mixing the marble powder with polymer component A and B, each are stored separately in sealed drums. These mixed material can be stored up to a month reliably provided the powder is without moisture when mixed.
When printing on site we utilised a professional two part polymer extruder. It was extruded a the ration 1:1 through high pressure pipes, attached to a custom mixing block which in turn connects the disposable static mixer nozzle. The nozzle is connected to the robot and curing in 2-3 mins, which can be accelerated chemically or by using heat.
The extruder is also mobile, considered “the fourth robot” it supplies material, power, controllers, heater control panel and extruder flow control panel. The fourth robot follows the small robots while printing demonstrating two levels of mobility within the system.
Notes:
The images included in this document may show differences due to the prototyping that was constantly evolving the system.
We have tried to make the robots as open as possible by sharing the specifications within this scientific paper.
http://arxiv.org/ftp/arxiv/papers/1406/1406.3400.p...
The project has also featured in a number of publications, see below for further information.
economist.com/blogs/economist-explains/2014/07/economist-explains-13
wired.com/2014/06/these-drones-could-3-d-print-your-next-house/
creativeapplications.net/environment/minibuilders-small-robots-x-large-sculptures/
Credits:
Research Team: Stuart Maggs, Dori Sadan, Cristina Nan, Jin Shihui
Faculty: Sasa Jokic, Petr Novikov
Project by: Institute of advanced architecture of Catalonia
Sponsored by: SD Ventures

Participated in the
Tech Contest
26 Comments
Question 1 year ago on Step 2
What kind of sensors for positioning where used?
7 years ago
"Small Robots"?!?
They really are tiny (compared to earth) :)
8 years ago on Introduction
What is the material that you are using?
8 years ago on Introduction
Brilliant work! I was hoping it was a more general print bot than the three separate specialized bots, but specialized bots could also be added to accomplish other tasks of construction such as plumbing and conduits for electrical systems. A small "pipebot" could make structural tubing around itself, possibly with a second print head for a rigid foam core. Coordinated little pipebots could make honeycomb cell structures for strength. A fiber placer could also be rigged to insert fibers that stick through and up out of the bead allowing higher strength before the need of the vacuum bot.
Man does this have my gears turning.
Thanks!
8 years ago on Introduction
Brilliant rethinking of 3d printing for large structures!
The flexibility of these robot's capabilities seem like they would be very impressive, but I find myself wondering if a much cheaper and simpler version could be designed to create simplified housing for disaster areas and other low income communities. I imagine something with a shape like a beehive sort of like the domes created by spirally winding continuous sandbags like these:
http://www.earthbagbuilding.com/projects/smalldome...
The device would not even need much in the way of computing power. The extruder could be guided mechanically, Sort of like this:
http://jimsoven.blogspot.com/p/the-dome-part-i.htm...
Of course for this purpose using local materials would be best, but if the polymer/marble dust is cheap enough it would still be practical.
9 years ago on Introduction
Is this thing DIY? Are there plans available?
Maybe there is a link here, but I don't see it.
9 years ago on Introduction
I would like to suggest that, rather than using suction to hold the third (suction) robot in place, you take advantage of the inherent complexity of the structures that the grip robot can lay down, and have it "build in" a track for the suction robot to hold onto and follow. For instance, the grip robot could make a small series of indentations beginning every 0.2 meters horizontally and continuing throughout that horizontal layer. Then it would create an overhang in subsequent horizontal layers. The third (suction) robot would have notched wheels that would fit into this track, holding it to the surface. This would eliminate the need for a complex, failure-prone suction system, and also provide a built-in self-metering (measuring) system that would improve position accuracy. If I'm not being clear please contact me and I'll make a diagram.
9 years ago on Introduction
that must be quite expensive!! how long did that take?
9 years ago
My brain just melted. Really quite awesome.
9 years ago
This is by far the most incredible and amazing thing ever to come out of instructables. This is so inspiring.
9 years ago on Introduction
How much for a boat? great ible!
9 years ago on Step 5
This is really cool. Basically a six-axis 3D printer. I want one.
9 years ago
absolutely awesome! 3D printed house anyone!??
Reply 9 years ago on Introduction
Oh yeah.
Waste-free construction is the future!
9 years ago on Introduction
Amazing work. Thank you.
And now for something that will make you laugh.
Your incredible project reminded me of part of a 1974 movie (not very good, except in parts) called The Groove Tube.
The part you have to see is this
9 years ago on Introduction
I take it the building material doesn't shrink much, or cracks might form. Rebar, or mesh, or some sort of fiber worked into the material is the usual solution in cement construction. This appears to have none. Do you ever wish it had some back-up fiber worked into the form?
A really nice instructable. Thanks.
9 years ago on Introduction
Awesome, but please edit the audio so the "music" is 1/2 the volume. The people speaking have accents (to English speakers) and the obnoxious volume smothers their voices. And those of us who find this amazing would prefer to hear them, not the electronic buzz behind them.
loved it, I often thought what a hoot it would be to connect a pump/feeder to a fast set plaster/cement kinda like spray on cement. As I have no ability to make that I love that someone else did.
Bravissimo
9 years ago on Introduction
I love your 1st paragraph :
There has always been a close relationship between architecture and technology. However, recently architecture has stagnated and the construction industry has been slow to adopt technologies that are already well established in other fields. Whilst we design digitally we still construct manually.
Thanks for sharing
9 years ago
That is extremely cool. Except for one picture in which the device is being pushed by someone, it is very hard to understand the size of the objects. A human operator perhaps for proportion in the images?
A time lapse video of this project will win the interwebz :)
Reply 9 years ago on Introduction
Just saw the video which was not visible in the instructable app. The process is well thought of and the robots work well. I especially liked the third one which sticks with suction.