Post-CAD Fabrication

From CS294-84 Spring 2013
Jump to: navigation, search

David Burnett - 4/2/2013 17:37:21

"Interactive Construction: Interactive Fabrication of Functional Mechanical Devices"

This paper allows designers to use a laser cutter directly by using a handheld laser pointer to trace shapes on a work piece. The piece is immediately cut by the cutting laser. Generating 2D designs in Illustrator or other such programs takes enough time that, for simply a prototype, such a drawing exercise may not be necessary. The system is reminiscent of the Sutherland light pen system in that it takes an approximate laser pen sketch and corrects/straightens/rights the strokes as appropriate. The major downside of this system is a lack of preview. The user must simply hope the correct stroke has been recorded and righted when the laser cuts the work piece immediately after. It would be simple to add a projector overhead to show what's to be cut, and I don't understand why this wasn't added. The lack of preview makes the more advanced laser tools like "butterfly" and "extrude" without a perfect idea of how the computer is interpreting laser commands and a perfect spatial memory.

"kidCAD: Digitally Remixing Toys Through Tangible Tools"

kidCAD allows young digital modelers to use a malleable gel to provide a computer with shape input instead of mouse and keyboard, which often require significant tranining. Kids can easily assemble a 2.5D model in the gel with a variety of shapes through purely kinesthetic means with video overlay feedback, making the barrier to entry for this system almost nonexistant. User studies for this work were performed in depth, revealing ways children may mispercieve the system and classifying major types of users. In the future. more specific systems may be built for each of the user types. The most significant limitation on such a hands-on system is the unavailabiilty of hands-on interactivity with the generated 3D models. Children may be unable to grasp the concept of a virtual image as something that can be modified if it's intangible. Secondly, the system is firmly 2.5D and can't image overlapping objects, which prevents full capture of what's created in the gel. Lastly, the paper prefers a particular brand of 3D printer for its color capability, but later notes color isn't captured by the system. A simpler, cheaper printer could therefore be used.

"Position-Correcting Tools for 2D Digital Fabrication"

This paper describes a way to augment 2D cutting tools such that their accuracy is significantly greater than that of the operator, be it human or servo. More than just a way to leverage gross movement to fine, this work automatically performs fine cutting operations within the vicinity of the larger operated tool. This allows CNC-quality production at a fraction of the cost and complexity. It's also adaptable to many materials and types of cutters, though the latter requires another bout of development for each cutter. The system's localization accuracy is dependent on a vision system, which means materials must be specially patterned. These patterns are undesirable in the finished product, so care must be taken to ensure they aren't affixed so weakly that they slip and cause the vision system to think the piece is stretching. There are also the usual downsides of a human-operated device: some users may move the tool too fast or slow and cause damage, and some may use the device improperly and ruin the piece or themselves.

arie - 4/8/2013 21:58:11

Constructables attempts to reconnect the user with his work piece by

 2 introducing a set of proxy lasers which allow the designer to interact
 3 with the raw material under a specified set of constraints.
 4 I find one of the problems of laser cutting is their one-way behavior - once
 5 you cut something and take it off the board, it is hard to realign the piece
 6 for the next stage, if additional cutting is required.
 7 I wonder if the proposed paradigm is intuitive to use as the feedback
 8 mechanism seems somewhat imprecise compared to a regular full featured CAD
 9 environment. Overall, while the idea of making the actual physical
10 fabrication process more interactive is very appealing, i think that the
11 implementation somewhat cumbersome.
13 KidCAD allows 2.5D models to be created from physical toys and other small
14 objects using the deForm surface interface including IR pattern projector
15 and an IR camera. This work presents an interesting alternative to 3d
16 scanning, with an interesting angle targeted at kids. In general i think
17 the idea of testing new HCI devices on children might be useful as it allows
18 the researcher to test his concept with less biases involved.
20 The last work, position correcting tools presents an additional attempt to
21 bridge the gap between the CAD design stage and the physical fabrication
22 stage. By allowing the human operator to position the tool approximately,
23 a small manufacturing tool could be used to fabricate devices much larger
24 than the motion range of the tool. Computer Vision based markers are used
25 to localize the tool in the bigger motion range. The QR codes seem to
26 add complexity to the system. The visual feedback on the display is a
27 great addition. this work made me wonder about the optimal involvement of
28 the user in the physical fabrication stage : full automated cnc/cutting
29 vs. intermediate intervention at the expense of added flexibility...

Valkyrie Arline Savage, PhD - 4/8/2013 22:03:37

This week’s papers share an excitement about bringing fabrication power out of staged tools and into the hands of people. This is an important step, since it retains the context of the material work being done.

I really like the Constructable. They endeavor to have a direct manipulation working system for laser cutting, but they also introduce some pretty cool ideas like “undo” for physical stuff (using the “union” or “butterfly” tools). I actually think that this is just the tip of the iceberg here... what if we had undo for other rapid prototyping tools? The laser cutter (and the vinyl cutter, its little cousin.. described in the MIT paper) has the advantage that it actually is fast and can give immediate feedback. An “undo” in a 3D printing design tool would take hours. The features of the tool chosen here really informed the possibilities. Anyway, some shortcomings that I didn’t really see addressed included their visual tracking mechanism and how it would work on reflective surfaces. They say that they use reflection-free lighting to help with tool positioning, and they didn’t discuss that this sort of means they can only use wood as a material. I like the idea of ubicomp as computer-mediated design without screens.

KidCAD builds off of the way interesting deForm material. For the first time in a long time, we actually see a real user study in a paper! They tested their project with kids, and even promise further evaluation in the future. How exciting! This paper endeavors to bring some other computer metaphors into real-world design, in this case copy-paste. They discuss the relative merits of rapid prototyping vs. sub-millimeter precision, especially in the eyes of children who want to use this as a creative tool rather than as a SolidWorks. I liked their technique of doing an early formative study using PlayDoh instead of just jumping into research without understanding. I liked, too, that they were very honest about how they were surprised by many of the kids’ uses of their tool. I think that the ability to give up precision and allow exploration is going to be powerful as we try to bring rapid fabrication tools into the hands of consumers.

Finally, the position-correcting paper was pretty cool. They are basically creating a computer that can be wielded for cutting. This is cool! They didn’t do an especially good job of describing everything in their paper... for example, I wasn’t sure how they would find the marker tape when it was on an angle (since they seem to be doing only horizontal and vertical searches through pixels?). I do think there is power in the ideas they offer, however. One that I like a lot is the ability to “map” the fabrication material and overlay the design on the map. This has some of the same spirit as the Constructable, since you align things by eye rather than calibration. Getting rid of calibration is awesome.

elliot nahman - 4/8/2013 22:10:21

Interactive Construction: The authors built a system which allows the user to use a laser pointer as a pen and “draw” on the top glass of a laser cutter. The laser cutter follows the paths they create cutting or scoring. Different “pens” allow for different actions.

They authors cite that their system allows the user to work directly on the workpiece. This does not entirely seem true as the user is separated from their piece. Plus their system makes use of one type of tool held in one way, as opposed to other workman’s tool whose tools match specific human capabilities and use different grips as a result. Their process seems most akin to using a Wacom tablet on a computer, but in many ways, less advantageous. The Wacom, you still get tactile feedback when you touch your pen to the tablet surface, which indicates a click/draw just as it does with a pen. With Constructable, that is not true. You are completely relying on your sight to tell where the laser is hitting a target a foot away and separated through glass. You are not so much working on the workpiece, which implies all the tactile experiences that come with it, as relying completely on your vision/hand-eye coordination to draw, but without as much feedback of where your pen just was. I also question their undo tool. Although useful, the number of circumstances when it is possible to use seems limited. Their ergonomic claims also seem absurd, though they do acknowledge that it would be better if it was tilted like drafting tables.

Ultimately, they cite three advantages to CAD: Fast interaction, trial-and-error, and precision. I remain unconvinced that their system upholds these advantages while providing the advantage of working directly with the laser cutter. Especially since they didn't seem to do a user test to see if their tool was more successful that traditional methods.

kidCAD: kidCAD is a tool to aid children in create, remix, and interact with digital models. It is a surface which allows users to take 3D object and create 2D scans plus some depth into a computer. They can edit and remix them by interacting with the surface.

The authors were inspired by playdoh and designs created with impressions of other objects. In general, it seems that they did a good job recreating this interaction. The output looks somewhat jumbled, but does mimic the constraints and advantages of the medium they were mimicking. That said, the authors started from a point of 3D modeling. Their tool, and they say so, does not support 3D and is more like working in 2D space. As such, it does not seem to have much value add to the original playdoh interaction, except that a 3D print can be made of their relatively flat object. I also find it a strange choice to constrain the space to 1:1. It seems like one of the large advantages would be that you can scale new objects in your remixes and thereby allow for greater flexibility and exploration rather that just reusing an existing object.

Position-Correcting Tools for 2D Digital Fabrication The authors created a tool to correct for human error in hand tools such as routers. This allows the user to operate hand tools while maintaining high accuracy of the intended shape. Essentially, they are taking tools such as CNC machines or laser cutters and replacing the big movements with they do with clumsy humans.

This is pretty amazing. However, laying down the markers seems potentially fraught with error. How accurate do these marker strips need to be? It seems like it would be quite easy to not apply them perfectly straight, or not perfectly spaced. I also wonder about the system’s ability to recover should the user go far enough off the path that the machine cannot compensate. What happens then? Will it find its way back gracefully once it is back on the line? The authors also did not talk much about the digital side of their tool. How easy is this to use? In my mind, maybe because they used a router as an example, this is most useful for those not fully agile in a digital environment, so its usefulness is completely dependent on how idiot proof that digital environment is.

Ben Zhang, PhD - 4/9/2013 0:48:26

  • Constructable (Mueller)

The constructable system empowers user to "directly" control the laser cutter in an interactive fashion, so that feedback is instant and users can enjoy the process of creation. In contrast to many other interactive design tools, constructable preserves the precision by defining constraints on the manipulation of users. The implementation involves a low-power laser beam for user to sketch and the real powerful laser cutter to perform actions.

Ideas from CAD help the system be more complete and powerful. Their proxy laser tools are primarily adopting tools from CAD designs. The "undo" functionality for physical cut has to cope with the irreversibility so that authors designed "union" function. To better illustrate the interactive procedure, this paper goes through a toy example. Overall, this is a nice paper describing their system and the ideas behind.

One great thing of directly dealing with "physical" cut is the instant feedback -- you know you are doing good or bad for this step. However, the problem with physical objects is inherently the limited versatility. When you sketch, the cut remains. And another sketch might make it difficult to tell from each other. Also due to limited display, many CAD illustrations which can assist design are not available. In such context, I am considering if virtual reality could help. The paper mentions some work that used projector. They mentioned that "Letting go of it not only increased precision, but also made sure we directed users’ attention ... ", I don't quite personally understand this argument. You could still have the constraints to make it precise. The attention attraction problem can be mitigated by adjusting the intensity, maybe?

Another consideration of this work is still the precision. [Disclaimer: I am not expert in CAD; never!] I usually have this impression -- before any serious construction, I tend to measure a lot and having those numbers in mind helps me design. Also in AutoCAD, you can specify numbers for any shapes you need. But this function is not provided in constructable. I understand this does increase complexity, but how necessary is it?

  • kidCAD (Follmer)

kidCAD has its clear target audience, and build the system primarily for them. The major point behind it is to combine tangible creation with digital capability to provide rich manipulation. Similar ideas have been seen in other systems like sifteo, where the creative use of many components form arbitrarily interesting system. kidCAD explores another space, where deformable gel helps children to create "CAD" files. The technical parts in the paper is the discussion of using IR projection and camera for implementation; while most of other contents focus on the design motivation and discussion around an empirical study.

Generally, the kids found the system interesting to use, and the study showed some general theme that how children tend to interact with the tool -- create patterns and textures, create pictorial scenes. They do combine several different input components, though most of the tasks can be categorized into stampers, sculptors and sketchers. The discussion with children support their initial design idea that having copying from physical world would initiate the interaction with the tool, to put with their words "when copying is made easy, copying becomes creation".

I like the way this work approaches. With the ever increasing number of commercial laptops/tablets, children are more involved in interacting with the virtual world. Though I don't have direct experience with the new generations, I do personally buy the argument that the trend is "pushing children away from physical and manual creation". Apparently, there will be the sacrifice of flexibility when CAD is put under a physical context. But when the targeting audience is children who aren't proficient in modeling precisely in a commercial CAD tool, it seems to make more sense to focus on direct intractability. And the unofficial study in this paper showed that most children like it.

  • Position-Correcting (Rivers)

The design of this position correction tool is well motivated. Though laser cutting and 3D printing are being largely adopted for fast prototyping, their inherent constraints on the size make people seek work-arounds such as glue and composition. With the help of users, this limitation can be overcome; however, careful design must compensate for the imprecision of human's operation.

The paper proposes a vision based technique to localize the device and auto-correct user's trajectory to match predefined path. Important characteristics of such tool include accuracy, robustness and low latency, and marker-based vision detection satisfy most of the requirements (especially accuracy, from the empirical study, an average error of 0.009"). The paper then discusses the details of implementation -- how the markers are designed, how the image detection is done, and how the actuation of motor works. A simple user interface which displays the relative position helps user to follow their plan, and the sample results ( map of United States) are pretty intriguing.

As far as I can recall, in one of Swarmlab seminar, Amit from MIT talks about a similar work which focuses on 3D creation. The benefit of combining human and computer over a purely computer-based approach is apparent. Humans are so agile and creative, while precision can be compensated by the adoption of computers. Though the concern of efficiency exists, the joy of creation is accompanying the process, making it never bored (unless duplication is needed).

Joey Greenspun - 4/9/2013 8:54:29

kidCAD: Digitally Remixing Toys Through Tangible Objects

These researchers developed a tool that uses physical objects such as a doll or an action figure to create a 2.5D digital replica via pressing the object into a deformable gel. They have targeted children and looked to make a more intuitive and easy means for them to be introduced to the world of digital design. I think this is a very unique niche the have targeted here. Usually when you think about digital design, you are trying to find a way to be more precise of more user friendly or more cutting edge, but these researchers aren’t doing any of that. In fact, they even agree to the fact that they have lost a bit it terms of precision, noting that it is not an immediate goal of theirs. Additionally, I really enjoyed all of the thought they put into making this a system that will help children understand the art of design and start to think about it in new ways. Being able to 3D print their creation after they are finished is brilliant. Being able to start with a doll, give it a dinosaur head, draw wings onto it, and then print it out in 3D is going to do wonders to the creativity of a child. Sure, just drawing it and seeing it on the computer might be marginally beneficial, but once the child can hold his/her creation and actually use it, that’s where the true awe is going to set in for that kid. Something I didn’t like, and I’m sure the kids were saddened by as well, is the lack of that last .5D. It wouldn’t be too hard for them to incorporate some sort of stitching function wherein the user could stamp both sides of the object and stitch it together in 3D. As a side note, the line “However, after Kindergarten, in western culture, children are encouraged to ignore play and the physical world in education [25]” seems terribly unjustified. And I’d love to read that source to see how they can stand to make that overarching claim.

Position-Correcting Tool for 2D Digital Fabrication These researchers have developed a tool to precisely cut out a complex shape programmatically, while only necessitating that a user moves a tool within a decently large range from the contour. This could all be done completely programmatically, however stages that can do things like this increase in price as their size increases. So, keeping the stage as small as possible is key. Their device does exactly that. By utilizing computer vision coupled with an actuating unit, they can cut out precise lines very quickly. A few problems I had with the paper are the need to first scan over the entire surface of the material using the tool, before any cutting is done. I wonder if you could, take a picture with a smart phone, or image the surface using some other technique and send that image to the tool. This scanning seems like it’d take a long time and could be easily changed. Additionally, the researchers just make a blanket claim that a user can keep the tool to within 1/8” of the contour of a shape manually. However, they don’t back this up with any field study or research. They need to justify this claim with some sort of user study. And along that same line, they should include some user study on the back end to see how effective their tool really is. In general, I do think this is a great idea and a pretty neat implementation. I would like to see them do some sort of line tracking so that one could just draw onto a piece of wood and have the machine follow the line that was drawn. The need to add the stickers/markers seems like it could be cut out if we assume a continuous line that the user wants to cut along.

Constructable This group focused on keeping the design and the end product in the same space. They mention that there is this disconnect when trying to model something in CAD and then laser cutting it in a separate space. Their goal was to make this design and real world cutting and modification happen in the same space. They achieved this using a system of low power laser pointers that all had different functions. They could be used for making gears, lines, circles, extrusions, etc. Each different laser pointer is aimed at the surface of the material in the laser cutter housing and the laser cutter performs the task asked of it by the user. This is achieved by using a specific button on the laser pointer, and the illuminated spot is tracked in real time by the system. This system has visual feedback, and yet does not use projection. They unit scores the object until the user is ready to turn the line into an actual cut. When I first started reading this paper, I was very skeptical about how this was going to work. I thought that everything was going to function basically as a free hand, tell-the-laser-where-it-should-cut sort of functionality. However, I was relieved to see how precise and programmatically things could be done. Even functions like scaling could all be performed using various laser pointers. I thought it was very beneficial that they modeled what they wanted their device to do based on what is so great about designing on the computer itself. Such as the ability to copy and paste, do things quickly, and have precise angles and shapes. They managed to accomplish keeping all of these beneficial design parameters in place, while also allowing the freedom of working on the material itself and on top of that, the ability to draw completely free hand. The ability to just see the beam as a reference (the middle button on the device) and not have it actually cut the object of interest is important for the users to be able to use this device as the researchers intentioned it. This is a beautiful hybrid of digital programmatic design and the freedom of free hand modifications.