The Self-Reconfiguring Robotic Molecule
A self-reconfiguring robot consists of a set of identical modules that can dynamically and autonomously reconfigure in a variety of shapes, to best fit the terrain, environment, and task. Self-reconfiguration leads to versatile robots that can support multiple modalities of locomotion and manipulation. For example, a self-reconfiguring robot can aggregate as a snake to traverse a tunnel and then reconfigure as a six-legged robot to traverse rough terrain, such as a lunar surface, and change shape again to climb stairs and enter a building.
We have designed a small robotic module we call the Molecule capable of self-reconfiguration in three-dimensional space. The Molecule (see figure below) is capable of independent movement on a substrate of identical Molecules, including straight-line traversal and 90 degree convex and concave transitions to adjacent surfaces.
A Molecule robot consists of two atoms linked by a rigid connection called a bond. Each atom has five inter-Molecule connection points and two degrees of freedom. One degree of freedom allows the atom to rotate 180 degrees relative to its bond connection, and the other degree of freedom allows the atom (thus the entire Molecule) to rotate relative 180 degrees relative to one of the inter-Molecule connectors at a right angle to the bond connection.
This movie shows how Molecules can reconfigure to change the shape of the robot. In this case, the task is locomotion--utilizing individual module motions, the structure can move itself. With only four modules the movement is limited to the plane, but with additional modules the system can perform more sophisticated tasks such as stair climbing or building towers to access locations out of the plane. File:walk4en-6 divx.avi
Contents for This Page
So far, there have been three hardware versions of the Molecule robot. The overall design, as described above, hasn't changed since the first version. Most of the changes have been related to the connection mechanism.
Version2a: The main difference between version 1 and version 2a was size. Version 2a reduced the atom diameter to 4" instead of the 5" atom diameter used in version 1. It was intended that version 2a would use the same 1" electromagnets as those used in version 1 but the 16V power requirement of the electromagnets was a design obstacle. Therefore, a working prototype of version 2a was never built.
Version 2b: Version 2b replaced the 1" electromagnet-based connector used in version 1 with 4 smaller electromagnets:
The small electromagnets only required 5V, making them more practical. However, the quad-electromagnet connector was much larger in diameter than the previous design, making some movements difficult. Another problem with the quad design was the difficulty of getting all the electromagnets to be coplanar. If the magnets are not coplanar, the holding force is greatly diminished. In the initial prototype, the electromagnets were well aligned, but it proved difficult to get the same performance on a second prototype.
Version 2c: Version 2c saw a return of the 1" electromagnet--this time a custom-designed version which could operate at 6V. Unfortunately, the holding power of the new design was inferior to the original 16V model (a higher voltage can achieve the same power with less current, allowing the winding wire to be thinner which allows more turns to fit inside the magnet). A working prototype of this design was produced, but did not perform as well as the first version 2b unit.
The difficulty with the connectors in version 2 forced a radical design change in version 3. Electromagnets were not working very well and their power consumption was also of concern. Because electromagnets must remain energized to maintain a connection, a great deal of power would need to be consumed to keep a structure from falling apart. This issue was realized from the beginning, but a practical alternative to electromagnets was proving difficult to find (we always assumed we would dump the electromagnetic connectors as soon as we found a better solution, but it wasn't happening). The key constraints of Molecule connector design are:
- Must be quad-symmetric (units need to attach to each other at at 90-degree orientation),
- Must be no taller than 23/32",
- Should consume power only when a connection is being made or broken,
- Should retract inside the atom bounding sphere when not connected, and
- Should be unisex.
This is a huge design challenge, especially the space requirement which leaves very little room for moving parts. Our analysis showed that it would be almost impossible to make an active connector within the 23/32" restriction, let alone one that could retract. However, we eventually realized that it might be possible to fit one active and one passive connector in 46/32" space. The only problem was constraint 5. This constraint was actually a leftover from our early design meetings--we had since realized that Molecules partition 3-D such that there are two distinct "flavors" of Molecules and each flavor can only connect to the other. Therefore, constraint #5 was not really necessary, but it seemed practical at the time it was conceived. The result is that we have abandoned unisex Molecules in version 3--Molecules now come in male and female varieties:
These movies show how the Molecule moves and connects to other modules:
Here is a movie of our first multi-Molecule reconfiguration: File:pair-i.avi Photo of a a four-Molecule structure:
In the following movie, the leftmost Molecule pair translates to the right of the 4-Molecule structure. The ability to translate a pair from one side of the structure to the other allows a global translation of the structure by successive pair translations. For a structure of this size it is necessary to provide a stabilizing base to which the male Molecules can attach. This is needed because there are not enough stationary Molecules to provide a stable structure when certain moves are done by a moving male Molecule (the torque needed to move a male can tip the structure made up of the other three non-moving units). The base would not be necessary if we had more units. Note that the speed of this movie has been increased by a factor of three--normally a gripper connection takes about 15 seconds which makes the movie a bit too tedious to watch in real time... File:translate2s divx.avi Despite what I said above, it is possible to do a pair translation without the base. This requires some non-standard angle rotations and some extra moves to stabilize the structure, but it works pretty well. As usual, this speed of this movie has been increased by a factor of three... File:translate2ens divx.avi This movie shows a complete locomotion sequence with each Molecule moving once. Again, it is performed without the stabilizing base. The speed of this movie has been increased by a factor of 6... File:walk4en-6 divx.avi Four high resolution snapshots of a reconfiguration sequence (1MB each):
The Molecule is controlled by two types of software: low-level assembly code in the onboard processor(s), and high-level code on a workstation. More on this to come...
Simulation is an important part of our Molecule research. It is difficult and time consuming to build hundreds of Molecule robots, however using simulation we can learn about systems with many more robots than we could build by hand. Our early simulations were simple systems of two or four units that demonstrated basic motion properties:
Our later simulations involved more Molecule modules. In these simulations we investigated walking, tower building, and stair climbing. This movie shows 20 Molecule modules climbing stairs: File:climb20cc-r.avi Our newest simulations use over 150 Molecules to show our scaffold-based planning method in action. Scaffold-based planning used tiles of Molecules (groups of 54 Molecules arranged in a specific shape) to create structures with tunnels in the interior. These tunnels can be used to increase the amount of parallel movement during a reconfiguration. The following movie shows a tile reconfiguration in which the tile on the lower right moves through the two center tiles to a position on the upper left: File:tile30-r.avi