Self-assembling robots take shape

From OE Reports Number 193 - January 2000
01 January 2000
By Sunny Bains

At Xerox PaloAlto Research Center (PARC; Palo Alto, CA), researchers are working toward the robotic equivalent of the Lego brick: a robotic block that, when combined with many others, can make complex, active shapes. Though the work in modular robotics has been in progress for several years, efforts have recently turned to extending the concept to true 3D self assembly. If successful, a heap of seemingly inanimate objects may one day be able to morph themselves into a chair, a cup (Figure 1), or a scuttling spider-bot.

Figure 1. Concept of the self-assembling robotic system; modules move to adopt a specified goal shape.

The reasons that Mark Yim and his colleagues are interested in building modular robots are clear. First, there is economy of scale. Mass producing tiny robotic blocks is much cheaper than designing, fabricating, and building custom robots -- especially if no one robot can be ideally equipped to do the required tasks. The composite robots, on the other hand, can potentially take whatever form is necessary to get the job done. In addition, they represent the ultimate in redundancy. Two broken robots could transform themselves into one completely "healthy" machine. Broken sensors can be fixed by simply going to the cupboard to get the right replacement brick.

Figure 2. A modular robot with cubic components.

Such a highly flexible system is not easy to design: there are serious hardware and software problems to overcome. When the work started as Mark Yim's PhD work in the early '90s, the blocks used were fairly limited in the ways they could be configured. Each module was essentially a cube that had two faces that were open, two that could bend (to provide actuation), and two that were fixed and could be used to join one module to another. Later modules were similar, and though they could perform various different types of locomotion, they worked best at forming simple shapes with essentially planar geometries (Figure 2).

The concept for the more recent work is shown in Figure 3. Here, the building block is a rhombic dodecahedron (RD). Its advantage is that these shapes can roll across each other (going from two faces in contact to the next two in contact) by being able to execute just one kind of move. This is easier to think of in terms of a hexagon, which Yim describes as the RD's 2D equivalent. Only three hexagons can meet at a single point. If three are actually present, no rolling around that point is possible because the space is filled. If there is just one hexagon, the shape has nothing to roll against. Only when there are two hexagons present can one roll around the meeting point, and the movement is always a 120-deg. turn. In addition, as for hexagons, it is difficult for RDs to slide against each other; this makes them a more stable choice than cubes.

Figure 3. A simulation using rhombic dodecahedron models to adopt a cup shape. Wire-frame elements represent modules not in their goal state. Filled elements show those that have reached their destination.

Once they determined the shape and movement of their building blocks, the Xerox team found a problem. Certain configurations that demanded large numbers of close-packed bricks were impossible either to get into or get out of (like a parking space that is exactly the length of the car). This was solved by allowing the vertices of the RD blocks to bend, thus allowing them to negotiate tight spaces.

Finally, researchers had to figure out a way to allow the distributed intelligence of the modular robots to quickly create shapes that were specified by a controller. Though the algorithm is complex, the basic ideas behind it are straightforward. First, goal destinations for the modules are defined based on the desired shape. These are prioritized going across the shape, so that activity happens at one end first and then moves to the other end (reducing the complexity of the problem). Modules then reserve spaces that they are closest to and that haven't been reserved by others, and proceed to them. All the reserving of spaces and the allocation of priorities is carried out through nearest-neighbor communication.

Simulations have shown that many shapes can be "grown" this way successfully, though not all come out perfectly every time. The next step is to finish the design and fabrication of the module hardware.


1. Mark Yim, John Lamping, Eric Mao, J. Geoffrey Chase, Rhombic dodecahedron shape for self-assembling robots, Xerox PARC SPL TechReport P9710777.

Sunny Bains
Sunny Bains is a writer and scientist based in the San Francisco Bay area.

Recent News
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?