SCALE System Overview

The SCALE (Self Constructing Auxiliary Living Environment) System is the outcome of Manuel Rueda Iragorri’s thesis for the Parsons MFA Design and Technology program.
In an attempt to accelerate the construction of proper living spaces for disaster victims by eliminating the qualified labor factor, this project explores the possibility of using flat-packed, solid panel housing modules and a robot that assembles them. The robot can navigate through a predetermined work area with the help of an overhead camera and an external computer, and deploy housing modules wherever they are needed to set up a refugee camp fit for human needs in a matter of hours instead of days.
As a way to summarize the entire project up to this point, a video to explain the details of the project as fast and accurately as possible and shots of the current robot and module are up next.


robotSidemoduleCollapsedmoduleExpandedsetupOverviewsetupOverview2

Assembly Sequence Videos

After multiple tries, a fully redesigned front for the robot, a couple of materials tested for the module and some fine tweaking, the system is finally operational. Up next is the video of the full assembly of the emergency housing module as performed by the drone (real time and fast versions).

Drone 2.0 update

For the past weeks I’ve been working on the physical and mechanical aspects of the construction drone. Most of the systems are working (the drive train, the slider, the drill and the compass work properly but the forklift still needs some tweaking) and the final physical form is being achieved. Up next are some pictures of the prototype at it’s current stage and a video of it moving around.
droneCurrent00droneCurrent01droneCurrent02droneCurrent03droneCurrent04

The full functionality should be done by next sunday and then it will be time for the navigation algorithm.

A Compass to Navigate

In parallel to building the new drone and module, I’ve also been working on improving the self navigation program for the drone. My approach for it’s improvement was to shift from presence detection to color detection, and that way I could easily determine which way was forward. Even though that approach seemed to work pretty well, I started finding out issues that I was going to bump into later, especially with perspective and ambient light changes.

colorDetect

I played around with different colored corners and front and back markers and in general terms, it works great, BUT that is assuming that the camera will always be positioned in such a way that the floor image will be close to bi-dimensional, which in real life will probably not be the case. For that reason, I consulted with Joel Murphy for potential navigation solutions and he suggested using an electronic compass module on the drone, that way I would always know which way is the drone facing.
I took Joel’s advice and got my hands on an HMC6352 compass module from parallax, and proceeded to make it work with Arduino. The module was fairly easy to use but it had a lot of call and response protocols within its programming, so I wrote a small library for the HMC6352 compass module that cleaned up a lot of what was going on in there. I then got the module to communicate straight to Open Frameworks, so now I’m ready to optimize my previous presence detection script and make it work with this new hardware add-on.
hmc6352CompassModuleWiringDiagram

Progress on Drone 2.0

Scaling the housing module was a very straight forward task, but the drone is a whole different ball game. On one hand the electronics (everything other than the power supply and the motors) remained pretty much the same, which means that there’s a lot more space to work with, but the mechanical components changed significantly. With the new weight of the components and the lessons learned from the previous drone, I had to add some new mechanisms to make the workflow more fluent and improve the functionality for a larger scale.
My first step was to design a box that would act as the robot chassis and the housing for the lead-acid battery, the motors and the circuitry. I decided to make it in 1/4″ acrylic because it was tough enough for the current prototype and could be easily worked with a laser cutter. The design was extremely simple and consisted basically of a slot for the battery and precision cuts for the axels of the motors. It originally included fittings for a full frontal axel but that changed for reasons that will be discussed in a paragraph below.

drone01drone02

Once the original design was built, I tested it and immediately bumped into a problem: the robot has no problem moving back and forth, but the friction generated by the wheels and the ground when the drone tries to turn left or right is too strong for the motors. Since the previous prototype was moving on treads, this situation was never a problem, but now I had decided to move to wheels because it would simplify manufacture at a larger scale and it would make the current prototype cheaper. In an attempt to correct that situation, I eliminated the front axle and attached 2 more motors instead. However, the experiment didn’t work. It barely shifted and that meant that once I added the full weight load, it would not shift at all. Given the time constraint, I made an executive decision and replaced the front axle for a rotating axle wheel (think shopping cart) so that the friction would be eliminated to a minimum when the drone was turning. Even though this solution wouldn’t be suitable for a real, open space scenario, I decided to go with it because it is more important to demonstrate the functionality of the system as a whole than showing that a drive train can work in unpaved terrain. ATV’s already exist, it’s just a matter of transforming the drive train in a further iteration.

drone03drone04

Once the modifications were ready, I proceeded to test the drone again and it worked smoothly. However, the motors still seem to be a little sluggish, so I ordered some right angle gear motors that are a lot stronger and will overpower any amount of weight I put on them at this scale.

With the chassis ready and tested, I moved on to design the rest of the mechanical attributes of the new drone. The new mechanisms include a lift for carrying the modules and a slider rail over which the drill will be mounted so that the housing modules can be carried closer to the center of gravity of the drone and still be assembled while not touching the ground. The final assembled mechanisms will look somewhat like the model below. There’s still a lot of work to do and not a lot of time, so I’ll keep working at twice the speed.

drone00

Test Circuit

One of the things that I really wanted to improve from the previous drone was the fact that it had 3 different power supplies that needed to be taken care of. For the new drone, I wanted to be able to use a single but very powerful battery, whose power I could regulate for whatever function I needed. For that reason I decided to go for a 12VDC, 12Ah, acid-lead battery. 12VDC because that is the rating of the motors and 12Ah because the motors consume a lot of current and there needed to be enough left over for the logic.

For the test circuit I just repeated the same assembly I had with the previous drone, but this time, each motor needed it’s own h-bridge (their individual current ratings added together exceeded the capacity of a single chip) and I didn’t need to piggyback them because the internal current resistance of the chips is imperceptible at this current level. below are the pictures of the test circuit and a video of it’s functionality.
testCircuitFartestCircuitClose

Scaling Up

In the past few weeks I’ve been doing two main things:

1. Prototyping the housing module at a larger scale (1:4 instead of the previous 1:20).

2. Material sourcing for both, the module and the construction drone.

I decided to work on a 1:4 scale because, first, it will allow me to spot structural inconveniences with the housing module second, because I couldn’t find any place suitable for me to be able to work at real scale and third because the prototyping price at this scale is still “reasonable” for my current resources.
In the beginning of february I started working on the scaled up module. The material of choice this time was foam core board because it would resemble the lightness of large scale panelling filled with insulation foam and it’s also remarkably fragile, so if there were any weak spots, they would be very easy to find. You can see the prototype in the images below.
scaledUp1scaledUp2scaledUp3scaledUp4scaledUp5scaledUp6

In this prototype I was able to observe something that will save me lots of headaches in the future. Some of the seams (the ones to both sides of the largest panels and the central seams of the roof panels) have a very large surface area and, in the smaller scale prototypes it was undetectable, but in this prototype I could clearly see that they will need to be structured with rigid elements. In the expansion process there’s no visible problem, but when the modules are to be compressed for storage, the larger seams have too much freedom, allowing the panels to “eat” them in random folds, creating deadlocks in the compression mechanism. To correct this I’m going to have to add rigid elements within the seams to reduce their freedom and guide them straight to where all materials need to be when the module is compressed.

Another interesting behavior of this module is that of the side doors. By having their own hinges, their expansion is not consistent, so for future prototypes I will have to lock them to the surrounding panels in order for them to work harmoniously with the others and help to strengthen the lower panels that might be compromised if they move by themselves and try to compensate the force of the other panels levering against them.

For the next prototype, I will also add the corner male/female locks so that the module can fit together with others like itself. The mechanism will be very simple (see image below) and will take advantage of the fact that the construction drone already counts with a lifting mechanism. In that scenario, the drone only has to run into a previously assembled module, lower the new one and release, and the two modules will fit together like a glove. In order for this mechanism to work properly, the front and side panels need to have 4 female hooks (one per corner) and the back panel will have 4 males. We’ll see more clearly how it works in a further iteration, and when it does, I plan to make the locks conductive so that electrical networks can be created by just putting the structures next to one another.
lockSketch

In terms of the material sourcing, I’m focusing in two main areas right now: finding key electrical/mechanical components for the drone and finding the right panelling material for the modules. In the electrical part, I already found some 12VDC geared motors that will be able to get the job done for the drive train, lift and drill mechanisms and a power supply strong enough to power them all and the microcontrollers that may be necessary. On the panelling side, I’m still inquiring around the five boroughs in order to find exactly what I need and for the right price, but that should be done by the middle of next week.
motorsAndBattery

My next steps will be to acquire the plastic for the module panels and the drone’s chassis, design the drone around the new components, test the power supply with the motors and the logic simultaneously and build both, the drone and the module’s new iterations as fast as possible.

The construction drone

constructionDrone

After determining the way in which I was going to expand and contract the housing modules, I moved forward and adapted the necessary mechanism to my previously made drone. Since I figured out a way to expand and contract the modules with a bolt, I needed some kind of drill adaptation (the one I made for testing in the previous post) that had to be leveled to the same height of the entry point of the modules. I proceeded to adapt a perforated board second floor (so to speak) to enable a firm base over my previously built circuit and attached the drill arm at the right height. I left some space open on the perf board for the circuitry that was needed to integrate the arm to the main board and moved on to connecting both systems.
circuitDiagramRelayMotorMy first step was taking out some leads from the main board and creating a plug out of them so that the assembly could be taken appart for maintenance if necessary. After having the leads out of the main board I started designing the circuit to control the motor from Arduino. Since the drill worked with a single 1.5V motor, using an h-bridge wasn’t possible because the current from a single AA battery wouldn’t be able to go through it, so I decided to use a 5V DC relay instead. I designed the whole circuit around the relay and it worked great on the breadboard, so I soldered the circuit and attached the arm to the drone and this is how it looks right now:
drillDrone1drillDrone2drillDrone3drillDrone4drillDrone5drillDrone6

Once the hardware was ready, it was time for testing and allowing something other than my hands to assemble my housing module for the first time. The following video was taken in that trial. The robot is driven manually from my laptop but there is no human intervention between the drone and the housing module throughout the whole video.

And here are some pictures of the process in case you want to take a closer look:
assembly1assembly2assembly3assembly4assembly5assembly6assembly7assembly8assembly9

The next steps for the project will be the following:
– Working on the navigation algorithm for the robot until it’s flawless.
– Redesigning the module as a tessellation element while preserving it’s mechanical atributes.
– Rebuilding the construction drone, not as a “Frankenstein” prototype but as an optimized product. Also add lift functionality to the drill arm so that the modules can be separated from the ground for transportation.
– Build the housing module at a larger scale to test its behavior.

One Bolt Assembly

For the last couple of weeks I had been crashing into walls figuring out how was I going to make the drone expand the housing module. My first approach was to enter through the bottom of the module through the opening I left underneath it in the new design and, with some sort of robotic arm, expand it. I started designing such arm and it started becoming more and more complex with every step I took. I started plotting my ideas first in free hand sketches and then in CAD software but the product I came up with didn’t fully convince me.mechanismSketchrobotArm01robotArm02robotArm04

The arm that I designed would pick up the module from the stack, put it in upright position, and expand it (by adding a couple of x structures on the top axle which are not accounted for in the images). However, the design started to look way too complicated for such a simple function and I started spotting problems everywhere. There was going to be a huge issue with leaving the modules in place and the multiple articulations would weaken the arm significantly to a point in which the timing belts and the pulleys would fail.
At this point I was running out of ideas so I consulted Marko Tandefelt for some guidance. I showed him what I had been working on and he agreed in that it was too complex. We had a long conversation with lots of sketching involved and ideas thrown out but he said two main things that hit me. First, rethink the stacking. I thought about the vertical stacking after dimensioning the module under Neufert’s architectural standards and comparing those values to the optimal numbers from the standard shipping containers, but, if people can modify the height of shipping containers to add refrigerating systems, why can’t I create a custom height for this product? So from that moment I started thinking about horizontal stacking as a possibility. The second thing Marko said was “your system should work like an automatic weapon”, meaning that it should be fail safe and things should always be in place when they are needed. When I was thinking about the vertical stack, the height of it would be changing every time I took one module out of the container, complicating the automation of a function. By doing a horizontal stack, leaves the same face of every module looking towards the same place every time I take one out, very much like the position of the primer in the bullets of an automatic weapon. Since the drone is already versatile on the x and y axes, I should try to work on those and avoid the z axis.
I sat down after our meeting and started looking for a way to open it from the side and then I saw the most simple solution of them all, a screw! By experimenting with previous models, I had already found a central spot in the upper part of the structure that allowed for the most even distribution of force throughout the module with the least amount of effort, so I drilled trough it, embedded a nut there and added a guide to the opposite panel to keep the screw straight. After the module was ready for experimentation, I took a 5″ long hex bolt and screwed it into the embedded nut and started seeing how the module started expanding accordingly. Success!
After confirming that the system worked, I wanted to mechanize it so I made an attachment for the drone with a 1.5 VDC motor, an old gearbox adapted from a toy, a DeWalt drill add-on for screwdriver bits and a custom joint made out of epoxy putty. The attachment created a drill head for the drone (I still haven’t attached it but that’s what it will do) and with the very low strength of a 1.5 VDC motor, it manages to build the housing module.moduleOpening01moduleOpening02moduleOpening03moduleOpening04moduleInsideView

The next step will be attaching the drill to the drone and linking the logic to make it work manually and automatically. Let’s see if I can do it in a week.

Remodeling the Module

After working on visualizing the housing module, I realized that there was a problem with the original design of the module. I hadn’t noticed the problem from the previous cardboard-made prototype because the material plus the tightness of the seams had made the module expand a little by itself, allowing some space in the middle. The fact that it was somewhat open, gave me the idea that I would have an entry point for my expansion mechanisms to come through and have some room to work.
collapsedModulefoldFlat
When I realized that there was no space to go through, I took a couple of steps back and rethought the design of the module so that it would allow the mechanisms inside of it for expansion. I sketched a potential model that would allow some space inside while at the same time be able to retain the previously adopted modifications for light and modularity.
module2.0Sketch
Once I was set on the new design, I moved on to building a model of it. This time around, I wasn’t going to make the same mistakes as the last time, so I built it out of acrylic panels and retained the duct tape as a simulation of the flexible membranes on the edges. I built the model at a 1:20 scale because it was appropriate for the size of the drive train that I made previously and paid close attention to the dilations between the panels. The result was very interesting.
module2.0.1module2.0.2module2.0.3module2.0.4

The structure seems to be sturdier and it expands and collapses easier than the previous one because the seams create less tension between the panels. The nodes created by the extra panels on the roof need special attention because there’s a corner shift when the module collapses, but that is easy to solve with extra membrane material or an elastic membrane, although a non-elastic has a longer lasting shelf life and it’s optimum performance is more reliable.
The 2.0 version of the housing module seems to allow more space for further evolution. The fact that there’s an access to the interior means that there’s a determined track to follow for the mechanisms that the robot might put into it and also enables me to play with the inner surfaces to create anchor points or railings to direct the motion of the robot.