Project Team: Dyani Robarge & Jack Fogel

Human Machine Virtuosity, Spring 2017

MoCap Carve | Final Video

ABSTRACT

The Augmented Carve seeks to develop a woodworking tool guidance system for aiding in the construction of 3D aggregated irregular geometries, in this case branches. The motion capture system tracks the location and orientation of both tool, a ryoba pull saw, as well as individual wood parts. A projection directs the user where to make half lap cut to join two beaches together, ultimately creating the structure designed on the computer through a series of members. 

OBJECTIVES

Rather than working with dimensional lumber, the half lap joinery connects a dense collection of large, irregular tree branches. Working with a catalog of highly-unique parts adds an extra challenge in building this augmented construction system. Our objective is to use this augmented system to make accurate, complex carves which vary in size, location and orientation. Our hope is that the precision of these complex intersections can be translated smoothly from digital model to physical work space. 

IMPLEMENTATION

To start, individual branches are made into high-resolution digital models through the use of open source photogrammetry software. Once the mesh models are properly scaled and cataloged, 3-4 nodes are placed on each digital branch model. Holes are drilled into the physical branches concurrently to ensure an accurate representation of the motion capture nodes in digital space. Once the nodes are fixed in place, the branch and ryoba saw can be linked and tracked as ‘rigid bodies’ within the motion capture software. Using a projector, tracking data is displayed on the work table next to the user.

The worker then affixes the branch securely to the table, making sure to keep the carve areas visible and accessible. Our team initially designed a fixture to affix each branch to the work table, but after several carve tests with branches of various sizes we quickly found the fixture to be an unnecessary element in the process. Once the worker has their branch securely fastened, the saw can be aligned to each cut area.

As the user moves their tool and a branch into the work cell, a dynamic user interface occurs. The camera zooms in onto individual carve regions and out to view the entire work space depending on the proximity of the tool to each carve region. Carve regions are marked by two outer alignment curves. The lines help the worker guide their tool in space. Circles above the saw endpoints act as a virtual level by giving clear indication of correct approach angles for each unique carve. Relaying real-time feedback of the tool’s movements through space allows the motion capture system to guide the user as they work.

OUTCOMES

We feel that this woodworking guidance tool provides intuitive feedback to the user and yields relatively accurate results, however there is room for much improvement. The low resolution of the projector greatly affected the tolerance of cut location, however we aimed to make most of the cuts smaller so with a small adjustment they would fix snug. The interface is very clear to us to use as we designed it, however its unclear if others would have the same ease of use as we did not test it with others. Overall, there are several features we noted throughout the process of creating this system which could be improved and expanded upon, but as the project stands it accomplishes much of what we had hoped.

CONTRIBUTION

Early in the project, Jack built a working prototype of the table clamp fixture. Dyani generated the overall assembly of branches and outlined carve regions for each. The two shared equal responsibility in the development of the overall workflow, its layout as a projected user interface, as well as prototyping different versions of the system’s Grasshopper script.

project context | motivation | scope | implementation

The system our team is developing harnesses the learned skill of woodcarving and applies this trade to an augmented assembly process. Specifically, the project seeks to augment the process by which individual branches are carefully carved.

VIDEO: Wood Carving Half Lap Joint with Hand Tools

The first step in this process is the tagging and reality capture of the branches. This process yields a digital representation of each physical branch used in the overall design. The clear resolution of each branch’s unique features in their digital representations allows the designer to mark each physical and digital model with three nodes. Each branch is set as a rigid body within the system. In addition, the carving saw is also set as a rigid body. Once this set-up is complete, the worker is ready to begin carving.

As a branch moves into the work cell, its tag appears on table’s projected screen. The worker then affixes the branch to the table’s fixture system, making sure to keep the carve areas visible and without obstruction. Once this branch is securely fastened, the wood carver introduces the saw to the work area and prepares to make the first alignment cuts. Lines projected onto the work table beside the wood carver help he or she align the tool in space and give clear indication of the correct saw angle when approaching each carve.

The system acts as a guide by relaying real-time feedback of the tool’s movements through space. By giving agency to the artist, they are free to decide when the visual cues are useful and when their attention is needed elsewhere in the work cell. This method of augmented working leaves complete freedom of the wood carve itself to the artist. We prefer the flexibility of augmented craftsmanship over automation because it allows for creative decisions to be made throughout the production process.

Our team plans to begin creating this system by setting up the table fixture and testing various motion capture markers on the branches. While prototyping, we will continue working on the code which links the work cell, branches and tool to the digital model. This abstracted visual of the augmented process is crucial to the overall workflow. We plan to refine the projected visuals so that the wood carver is able to work seamlessly while interpreting information needed to complete the piece.

We began this exploration by interviewing Miranda Miller, a CMU freshman art student who specializes in clay sculpting work.  From the information we gathered, we chose 3 tools that we thought were essential to creating a clay sculpture from start to finish.

Afterwards we invited Miranda into the dFAB lab and used the MOCAP software to capture her working.  To do this, we designed and lasercut structures to hold 3 small, reflective spheres which the software uses to track rigid bodies.  For the software to work, the bodies must be unique, asymmetric, and not inhibit tool motion. We also modeled each tool in Rhino.

We placed Miranda and her block of clay on one side of the table.  After looking at the data in MOCAP, we realized that we would drop fewer frames if she was facing the other side.  We recorded her sculpting with all three tools, and making Once we got the MOCAP data, we then analyzed the data using Grasshopper.  First, we combined the MOCAP CSV data and the Rhino tool models to create graphics of the tool moving throughout the space in Rhino.  We then decided to focus on the small tool and it’s movements, because the data was clearest and most interesting.

From that we have isolated several gestures that we found interesting and observed the velocity of the tip of the tool visually.  Within Grasshopper, we created a color gradient for the points that represents the velocity at each point.  We also exported the velocity matrix as a CSV and graphed it in MATLAB.

This data shows a trend in how an artist uses the small tool to smooth surfaces.  The tool is oriented more horizontally and moves much slower than normal back and forth motion.  This data can be used to help a new clay sculptor gain the skills of an expert much quicker.

We then made a digital model of the sculpting motion.

SculptMotion is a project by Lissa Biltz, Cecilia Ferrando, Atefeh Mhd, and Dyani Robarge.

Special Thanks to Miranda Miller, ’20   for all of her help.