The Research

This summer, I will be working with Dr. Jenny Walter to study metamorphic robots.

What are metamorphic robots?

Metamorphic robots are robots able to change their shape without outside help. The robots are composed of a collection of independently controlled robots that can move around on the other robots to reform.

The image below shows how a module moves. The module labeled S cannot move while another module is moving around it. The moving module wraps itself to another edge of the still module. Then it disconnects from the edge it started at and wraps itself back to hexagon shape. (Beth, the DMP student working on this project the two summers before this one, made this image.)

In our definition, every module has the identical structure, motion constraints, and computing capabilities. (All information needed by each module is computed by itself.) The modules also have a regular symmetry, so they can be packed without any gaps between them.

These robots that can change shape and move without outside intervention are useful in environments where people cannot go. Examples of such situations are out in space, in mines, deep underwater, and in burning buildings.

Having many identical modules makes the system more robust and more cost-efficient. If one module breaks down, the whole system can still continue. The modules can be mass-produced cutting down on manufacturing costs.

Giving the modules enough information to move without colliding with other modules is tricky. In the research I'm working on, the modules don't communicate with each other, and they don't have any way of sensing modules that are not attached to them. This makes it hard for the modules to determine where to move without colliding into one another. I'm working on an algorithm for each of the modules to run that will deterministically move each of the modules to a cell in the goal configuration when there are obstacles in the way. This means that each of the modules will know where to go without running into anyone else or the obstacle. The algorithm will work for any admissible, or possible, initial and goal configurations. Because it is deterministic, the algorithm will work 100% of the time. Some researchers working with metamorphic robots developed probabilistic algorithms; they only work most of the time.

Below are screen shots from the simulation. The goal cells, the cells we want to fill, are colored red. The obstacle cells are colored green. The goal cells inside the obstacle, the pocket cells, are colored light blue. The yellow cell is the substrate path (or the cells used to roll along while getting to where they want to be). The substrate path would be much longer if the obstacle wasn't so close to the west side of the goal configuration. There can also be a substrate path to the east of the obstacle if the goal cells extend far enough past the obstacle. The black cells are "repaired." That means they are cells east of the obstacle that need to be filled before the other cells to prevent deadlock. The blue cells are the modules in the initial configuration. Initially, the modules are only in the blue cells and then are trying to move to the red, light blue, yellow, and black cells.

Chirkjian has a movie of a real hexagonal metamorphic robot on his site.

Papers, papers, and more papers.
This is a list of papers on metamorphic robots I have read so far.

You can't research metamorphic robots without reading some Chirikjian.
"Bounds for Self-Reconfiguration of Metamorphic Robots" by Chirikjian and Pamecha.
"Kinematics of a Metamorphic Robotic System" by Chirikjian.
"A Useful Metric for Modular Robot Motion Planning" by Pamecha and Chirikjian.

Papers directly related to my topic:
"Distributed Reconfiguration of Metamorphic Robot Chains" by Walter, Welch, and Amato (my mentor and advisors).
"Concurrent Metamorphosis of Hexagonal Robot Chains into Simple Connected Configurations" by Walter, Welch, and Amato.
"Choosing Good Paths for Fast Distributed Reconfiguration of Hexagonal Metamorphic Robots" by Walter, Beth Tsai, and Amato.
"Enveloping Obstacles with Hexagonal Metamorphic Robots" by Walter, Tsai, and Amato.
"Concurrent Reconfiguration of Hexagonal Metamorphic Robots: Algorithms for Fast Execution and Obstacle Envelopment" by Walter, Tsai, and Amato.

Related papers are also good to read.
"Scalable Parallel Algorithm for Configuration Planning for Self-Reconfiguring Robots" by Kotay and Rus.
"Emergent Structures in Modular Self-Reconfigurable Robots" by Bojinov, Casal, and Hogg. I really enjoyed this paper.
"Rhombic Dodecahedron Shape for Self-Assembling Robots" by Yim, Lamping, Mao, and Chase.
"Controlled Module Density Helps Reconfiguration Planning" by Nguyen, Guibas, and Yim.
"A Complete, Local, and Parallel Reconfiguration Algorithm for Cube Style Modular Robots" by Vassilvitskii, Yim, and Suh.
"Self-Reconfiguration Planning with Compressible Unit Modules" by Rus and Vona.
"Self-Reconfiguration Planning for a Class of Modular Robots" by Casal and Yim.
"PolyBot: a Modular Reconfigurable Robot" by Yim, Duff, and Roufas.
"Snake Robot Free Climbing" by Nilsson.
"M-TRAN: Self-Reconfigurable Modular Robotic System" by Murata, Yoshida, Kamimuri, Kurokawa, Tomita, and Kokaji.