Integrating 3D Vision for Advanced Robotic Construction

01 Story

Vikas Enti, Aaron Small, and Felipe Polido share a common goal: to leave the world a better place than they found it. Reframe Systems is built on this ideal. The group of co-founders have over 20 years of combined experience at Amazon Robotics, bringing world-class expertise in how to automate and scale large operations. The company was born when Vikas partnered with Eclipse Ventures to develop a thesis on how to use the team’s expertise to make a major impact in curbing climate change. 

Today, the built environment accounts for upwards of 40% of global emissions worldwide. In the United States, there is an acute housing shortage, exacerbated by the high costs of construction and a skilled labor deficit. The upsized carbon impact, along with the high potential of industrialization, made residential construction the perfect sector for high-impact intervention.

After building conviction in the business case, the founders formed Reframe Systems with a seed investment from Eclipse Ventures and Foundamental Ventures. The team then brought their mastery of robotics, engineering, manufacturing, and operations together to create a design for a production network of flexible, robotic micro-factories.

The team at Reframe Systems are on a mission to unlock climate-friendly housing for all. Reframe is vertically integrated to perform the full spectrum of design & engineering, construction, and site work necessary to deliver turnkey homes to their customers.  Their primary focus is on missing middle housing typologies, bridging the gap between single-family homes and high-density apartments.

02 Challenges

In order to truly meet the US’s housing demand while ensuring a cleaner future, the home-building industry needs to produce one new net-zero home every 15 seconds. As it stands today, the industry is ill-equipped to meet this challenge, and industrialized construction should be a key to unlocking production in the face of rising costs and a dwindling labor force.  However, modular construction, as it stands, has major technological barriers to scalability. 

Not all US modular factories use automation.  Those American modular industry players who do leverage automation rely on third-party, off-the-shelf solutions for their factories. These systems are designed for high throughput but require a large initial investment and factory space. In addition to the high capital layout required for these factories, production lines are often limited in their design flexibility by both layout and the software that controls their automation.

With over 30,000 zoning jurisdictions in the US, there is a huge range of requirements across the US market that can put these rigid systems at a disadvantage.  The market needs a solution that can be deployed quickly, with minimal capital investment, and that has the flexibility to easily adapt to local requirements.

03 Solution

The solution comes in the form of Reframe Systems’ software-driven production system, which combines augmented manual workstations and automated building blocks.  Reframe’s production system is designed to operate in robotic micro-factories with low capital and operating expenditures so that production can be localized close to demand centers. At the work cell level, their flexible robotic framing cell has a footprint of only ~500ft^2 (~46m^2). The small footprint means the work cell can scale laterally for different factory layouts and desired throughput, providing operational flexibility.

Reframe’s vision is to leverage robotics to accelerate the automation pace of dirty, dull, and/or dangerous tasks so they can focus their build team on the tasks that benefit most from human dexterity.  Reframe’s robotic work cells can perform multiple types of tasks, including wall-panel framing and sheathing. Wall-panel framing is the process of assembling the structural elements inside walls. The sheathing is the process of affixing gypsum or plywood boards to the wall frames.  In the US residential market, these components are commonly constructed with dimensional lumber, a construction method also known as light-wood framing.

Front view of the Reframe System robot

Since lumber is a natural product, a builder must do a full quality assessment of the materials, checking for unique characteristics like twisting and bowing. For lumber quality control, you also need to be able to detect irregularities as small as 1/16” (1.6mm) along the length of the board such as staples and nails. In a typical, manually performed wall-framing process, carpenters sort for quality and unique properties and adjust accordingly.  To migrate this function from people to robotics, a system must be able to assess materials to the same level or better than the human eye. 

Reframe’s workcell is supported by the Zivid 2 L100 camera in combination with a patent-pending universal magnetic mounting surface. This magnetic assembling jig serves as a manufacturing blank canvas, and combined with Zivid’s vision capabilities, provides the foundation of a flexible and continually upgradable workcell. In the long term, this system will be able to pick and place lumber, panels, insulation, and glazing, apply nails and staples as required, and all while assessing materials quality before and after assembly.

Reframe Systems use a custom-made gripper and a patent-pending universal magnetic assembly jig

All materials are delivered to the work cells via operator-loaded pallets. The robot in the work cell can handle a diversity of presentation, orientation, and ordering using a vision system. All of the materials are labeled with QR markers that can be read by both camera and human so that they can be properly placed according to the design. 

The Zivid 2 3D camera increases the cell's functionality by providing both 2D and 3D vision to the solution. At the moment, the camera is used to identify lumber pieces via the QR markers on the lumber and then determine the robot's pick poses. Moving forward, the Zivid camera will be integral to unlocking the full suite of tasks performed by the robotic work cell.

The Zivid camera was key to developing and testing Reframe’s work cell.  The Zivid projector function has helped the team test pick points, reducing errors and helping developers test and verify. The projector function also helps to guide operators through visual cues, an innovation that has shown promising results.