3D printing typically involves a single machine creating an object. But a team of undergraduate researchers has taken that paradigm far outside its box in the Design, Research, and Education for Additive Manufacturing (DREAMS) Lab, led by L.S. Randolph Professor Chris Williams.
The project started with the goal of 3D printing a drone that could autonomously fly out of the printer and evolved into a highly robust, adaptable approach for manufacturing large mechatronic systems.
Drones are complex mechanisms with spinning propellers, precisely calibrated electronics, batteries, and an array of other pieces and parts. One of the team’s early challenges was approaching that complexity with 3D printing. Some parts, such as a charged battery and motor, simply couldn’t be printed and instead required assembly. This assembly doesn’t fit with the normal function of a 3D printer, which is to lay down stacked layers on top of one another, forming a single stationary object.
The team also needed to solve a somewhat sticky problem: setting the drone free. Because a 3D printer’s first layer slightly adheres to the build plate, the team needed to figure out how to scrape the printed piece off during the autonomous process.
A typical 3D printer wouldn’t achieve all these objectives. While the machine can place layers to create a shape and even print different materials together, it cannot grab components and assemble them, and it cannot scrape off its own product. Determining how to complete all those steps autonomously would be the critical factor in creating a working drone that successfully departed its maker.
The box becomes an arm
The team planned to circumvent the typical 3D printer’s function that uses a print head on a fixed gantry. Instead, the group used a robotic arm that could be equipped with tools for printing and component assembly. Although the arms take more work to program and operate, they also offer more options.
“By attaching a print head to multi-axis industrial robotic arms, we gain additional degrees of freedom of movement that allows us to finally print in true 3D,” said Williams. “Instead of just stacking a series of two-dimensional prints of each layer, the kinematic flexibility of the robot arm allows us to deposit material in any direction in 3D space.”
The robotic arm offered team members another advantage: They could make a versatile set of tools for the arm. This multimodal approach meant that the arm could use a 3D print head for one part of a job, change to an assembly tool for placing electronics and other finished parts, and then switch back to a 3D printing tool to close up the drone frame. By deploying multiple tools in a single robotic work cell, they removed the need for multiple machines.
“The flexibility of robotic arms allows us to change tools mid-print so that we can place foreign objects such as motors, batteries, and wires into the object while it is being printed,” Williams said. “This provided us with a path to fabricating complete functional mechatronic assemblies in a single robotic work cell.”
Breaking the bond
After the students designated parts for either printing or assembly, they needed to determine the best way to break the sticky bond between the finished piece and the build plate. This became a lesson in thermodynamics.
The robotic printer’s large, heated build plate, which acts as a base for the piece being built, is slightly warm to create enough adhesion so that the piece doesn’t move during printing. Though essential to achieving the precision required to build, it created an obstacle for a printed piece that needs to fly away from its printer.
Team member Dalton Phillips discovered a surprisingly simple solution to remove the printed piece: let it cool. If the plate cooled a few degrees after the print finished, the adherence became weaker. Once the plate cooled to a certain point, a simple mechanical scraper could be used to push it away. This approach worked, and the drone could be set free.
The plan comes together
With the fundamentals established, the team set about completing its objective: a drone that flew away from the printer. Over several months, team members created multimodal tools to be fitted onto their robot arm. Because they were using a robotic arm that the DREAMS Lab had previously designed for 3D printing, the printer machinery was readily available. The other tools needed for assembly would have to be created or modified to operate on the same arm.
What does a 3D printing team do to make modified parts? Use a 3D printer, of course.
The team built new fittings for the arm, adapting the machinery on the fly to make it work more efficiently and to correct problems as they arose. Team members were tackling the project from both sides: manufacturing custom machine parts to perform necessary tasks, and deploying those parts to enable autonomous manufacture of a fully functioning drone.
Beyond designing the hardware, the students spent countless hours writing programming. When the size of a drone chassis was wrong, they modified the size of the print. When an assembly tool couldn’t hit its mark to place electronics, they reprogrammed its movements to correct. If a drone was finished but crashed off the side of the printer when it powered up, they changed the drone design.
Hundreds of hours of trial and error were expended as they solved problems from multiple angles.
On April 25, they finally achieved their first successful print and flight. In a fully automated build, the chassis was printed, the electronics were placed, the rotors spun, and the drone flew away. Immediately after the drone’s departure, the robot went back to work to autonomously print an entirely different drone design with a new set of modular electronics.
The team was motivated by the success to pursue new innovation. One of the first objectives became improving the robot’s tools.
In building their first multimodal machine, the students had created a single tool for their robot that both 3D printed and placed modular pieces. While it was well-suited to hitting their early marks, it was not ideal. They had essentially created two hands on the end of an arm, so its ability to maneuver in tight spaces was somewhat restricted by its many appendages.
The team also wanted to equip far more than the two tasks enabled by that first tool. The wish list included tasks such as wire embedding, 3D scanning, and trimming and soldering. Students also wanted to add the flexibility to create and equip new tools for future builds.
The new objective became the ability to automatically exchange single, modular tools on and off so the robotic arm could get the exact tool it needed for each task. Unfortunately, the small robot used in the first run had limited reach, and placing a rack of tools in that space would create less real estate for it to do its work.
Students needed to change their robot.
They didn’t need to look very far to find a good solution. Postdoctoral researcher Joseph Kubalak, also a member of the DREAMS Lab team, had been working with a larger robotic arm capable of changing tools as the team required. The larger robot could also move around in an expanded work area, which gave the group the ability to set up a hefty arsenal of tools to attach. In the process, students also found new ways to streamline the process.
“Transitioning to the larger robot was not as simple as unplugging the old one and plugging in the new one,” said team member Kieran Beaumont. “Because many of our original hardware and software design choices had been made specifically for the smaller robot, almost all the tools, electronics, and code had to be redesigned. This experience taught us the importance of modular design, not just for the drones we were building, but the work cell too. Using what we had learned, we designed the new work cell to have tools with standard interfaces and power requirements and developed a control system capable of operating any size of robot arm.”
The team’s expanded work area also means a wider range of final products. In its first project, the size of the drones built was limited by the reach of the smaller arm. With the larger arm’s increased size and tool access, the team can construct much larger drones.
Beyond drones: Building a mechanical ballet
Having expanded the original idea to a bigger stage, the team’s next step is building a team of robots that work together. The students envision a future where 3D printing is combined with assembly for not only one robot but multiple robots working together. A team of mechanical arms that move among one another seamlessly, flowing together like a mechanical ballet, is their objective. This will require a new group of research collaborators, whom the team is currently pursuing.
“We have faculty who are really good at robotics, really good at making robots collaborate together,” said Kosmal. “We want to invite those people together to ask how robots can make really cool things.”
Those “really cool things” cover a host of possibilities. As the system currently operates, it could be used by NASA to automatically produce a host of different mission-specific drones deployed for remote work on Mars or enable on-demand drone fabrication for finding survivors and delivering supplies during disaster relief scenarios on Earth.
“I think this project speaks to the future of additive manufacturing,” said Williams. “It is time to move beyond printing static parts in premade boxes and time to start thinking of ways to integrate 3D printing technology into advanced manufacturing workflows to enable the creation of truly multifunctional products.”
The original team of undergraduates who tackled this project included Tadek Kosmal, Kieran Beaumont, Eric Link, Conner Pulling, Dalton Phillips, Heather Wotton, Camille Kudrna, James Lowe, and Hutch Peter. Many of the team members have since graduated, with several now working in industry. Kosmal, Beaumont, and Wotton are pursuing graduate degrees in the DREAMS Lab, and team member Link is doing the same in the lab of Kevin Kochersberger.
“This project has been a longstanding dream of mine,” said Williams. “Our lab has been working on some of the individual elements of this vision for quite some time, and seeing it all come together by this talented group of students – many of whom have been working in our lab since their freshman year – has been extremely rewarding and inspiring.”
This project received an initial award of $75,000 in funding from the NASA University Student Research Challenge, followed by an additional $40,000 combined awarding from Boeing, Braskem, Cube Pilot, KDE Direct, Northrop Grumman, RoboDK, Stäubli, and Xoar.