A fully autonomous, self-replicating robotic assembly system that can both assemble larger structures, including larger robots, and plan the best build sequence is years away, says Gershenfeld. But the new work makes important strides toward that goal, including working out the complexities of when to build more robots and how big they should be, as well as how to organize swarms of bots of different sizes to efficiently build a structure without too collide each other.
As in previous experiments, the new system involves large, usable structures made up of a series of tiny identical subunits called voxels (the volumetric equivalent of a 2-D pixel). But while previous voxels were purely mechanical structural parts, the team has now developed complex voxels, each capable of transporting both power and data from one unit to the next. This could enable the construction of structures that not only carry loads, but can also perform work such as lifting, moving and manipulating materials – including the voxels themselves.
“When we build these structures, you have to build in intelligence,” says Gershenfeld. While earlier versions of assembler bots were connected to their power sources and control systems by bundles of wires, “the idea of structural electronics – the fabrication of voxels that carry both power and data as well as power” – emerged. Looking at the new system in operation, he emphasizes: “There are no cables. There is only the structure.”
The robots themselves consist of a chain of several voxels that are strung together. These can grab another voxel via attachment points at one end and then move like a caterpillar to the desired position where the voxel can be attached to and detached from the growing structure.
Gershenfeld explains that while the earlier system demonstrated by members of his group could, in principle, build arbitrarily large structures, as the size of those structures relative to the size of the assembly robot reached a certain point, the process became increasingly inefficient, leading to longer and longer paths , which each bot would have to cover in order to bring each piece to its destination. At this point, with the new system, the bots could decide it was time to build a larger version of themselves that could reach greater distances and reduce travel time. An even larger structure might require another such step, with the new larger robots creating even larger ones, while parts of a structure containing a lot of fine detail might require more from the smallest robots.
As these robotic devices work to put something together, Abdel-Rahman says they face choices at every step along the way: “It could build a structure, or it could build another robot of the same size, or it could build a bigger robot.” ” Part of the work the researchers have focused on is developing the algorithms for such decision making.
“For example, if you want to build a cone or a hemisphere,” she says, “how do you start path planning and how do you break that shape up” into different areas for different bots to work on? The software they developed allows someone to input a shape and get an output showing where to place the first block and each one after, based on the distances to be covered.
According to Gershenfeld, thousands of articles have been published on route planning for robots. “But the step after that, the robot has to make the decision to build another robot or a different kind of robot — that’s new. There is really nothing before that.”
While the experimental system can handle the setup and includes the power and data connections, in the current versions the connections between the tiny sub-units are not strong enough to support the required loads. The team, which includes graduate student Miana Smith, is now focused on developing stronger connectors. “These robots can walk and place parts,” Gershenfeld says, “but we’re almost — but not quite — to the point where one of these robots will make another one and it’ll walk away. And that’s down to fine-tuning things like the power of actuators and the strength of joints. … But it’s gone far enough that these are the parts that will lead to it.”
Ultimately, such systems could be used to construct a variety of large, high value structures. For example, the way planes are currently built involves huge factories with portals that are much larger than the components they’re building, and then “if you’re building a jumbo jet, you need jumbo jets to power the.” Carrying parts of the jumbo jet to make it,” Gershenfeld says. With a system like this, built from tiny components assembled by tiny robots, “final assembly of the aircraft is the only assembly.”
Similarly, making a new car “can spend a year tooling” before actually building the first car, he says. The new system would bypass this whole process. Such potential efficiencies are why Gershenfeld and his students have worked closely with automotive companies, aerospace companies, and NASA. But the relatively technically undemanding construction industry could also potentially benefit.
While interest in 3D printed houses has increased, these now require printing machines that are at least as large as the house being built. Again, the potential for such structures to be assembled by swarms of tiny robots instead could offer advantages. And also the Defense Advanced Research Projects Agency is interested in working on the possibility of building structures for coastal protection against erosion and sea level rise.
Aaron Becker, associate professor of electrical and computer engineering at the University of Houston, who was not involved in this research, calls this paper “a home run – [offering] an innovative hardware system, a new way of scaling a swarm, and rigorous algorithms.”
Becker adds: “This paper examines a critical area of reconfigurable systems: how to quickly ramp up a robotic workforce and use it to efficiently assemble materials into a desired structure. … This is the first work I’ve seen that approaches the problem from a radically new perspective — using a rough set of robot parts to build a series of robots, sizes optimized to achieve the desired structure (and others robot) to build as quickly as possible.”
The research team also included MIT CBA student Benjamin Jenett and Christopher Cameron, who now works at the US Army Research Laboratory. The work was funded by NASA, the US Army Research Laboratory, and CBA consortia.