Lab Info
Journal Club
In The News
Upcoming Meetings
Distributed Robotics


Modular Robots Distributed Robotics Networked Robots Animals and Robots Sensor Networks Underwater Robotics 3D Navigation Desktop Robotics

Cellular Automata Adaptive Controllers Inverse Kinematics

Modular Self-Reconfiguring Robots

We wish to develop versatile and extensible massively-parallel distributed robotic systems. We believe that versatility can be obtained by developing self-reconfigurable systems. Self-reconfiguring robots have the ability to adapt to the operating environment and the required functionality by changing shape. They consist of a set of identical robotic modules that can autonomously and dynamically change their internal geometric structure for locomotion, manipulation, or sensing purposes, in order to optimally carry out a variety of tasks. For example, a self-reconfiguring robot system could self-organize as a snake shape to traverse a narrow tunnel and reorganize as a multi-legged walker upon exit to traverse rough terrain.

Self-reconfiguring robot systems employ physical mechanisms that allow unit modules to dynamically and automatically configure themselves into a more complex robot. Creating a self-reconfiguring robot systems poses many engineering challenges. These challenges are centered around designing the basic self-reconfiguring module and the inter-module connections, and aggregating distributed systems from these modules.

The Robotic Molecule

We have already designed the Robotic Molecule, a four-degree-of-freedom, small scale module that can aggregate with other identical modules to form three-dimensional dynamic structures. The initial goal for our design was to create a solid shape that could closely be packed in 3D space and that could move by attaching itself to similar units. We chose a design based on cubes. It consists of two atoms linked by a rigid 90 degree connection we call the bond. In the current design, the bond can extend and retract a small distance. Each atom is a cubic structure that has connectors in the middle of each face. There is one connector to the bond and five inter-molecule connectors. Each atom has two degrees of freedom: one degree of freedom allows rotation about one connector (the bottom connector); the second allows the atom to rotate about the bond. The rotation axes that are possible due to the two connector actuators are perpendicular to the rotation axes that are possible due to the bond actuators. Thus, the module is capable of in-plane and out-of-plane motions. It is worth noting that only three of the four degrees of freedom are essential for three-dimensional self-reconfiguration. The additional actuator provides symmetry (and thus easier control algorithms). This design enables a robotic molecule to connect to identical modules and move independently on a three-dimensional substrate of identical modules. This capability supports automatic shape transformations and the creation of dynamic structures that can move using self-reconfiguration

The current Projects for Self-reconfiguring Molecule Robots are:

  1. Creating distributed systems of Self-reconfiguring Molecule Robots
  2. Developing distributed algorithms for shape metamorphosis planning
  3. Developing algorithms that use self-reconfiguration for planning vestatile locomotion gaits; issues include mapping the terrain type to gait, keeping the static and dynamic stability of the shape, etc.
  4. Developing algorithms that use self-reconfiguration for self-repair
  5. Creating active sturfaces with Self-reconfiguring Molecule Robots
  6. Building MEMS self-reconfiguring robots

The Crystal Robot

An alternative design for self-reconfiguring robots we are currently building is called the Crystalline Module. This module has cubic shape with connectors to other modules in the middle of each face. Our idea is to build a cube that can shrink by a constant fraction (we hope to build a module that can be shrank by factor of two) and expand to the original size. The principal direction (e.g., x, y, z) will be shrank separately or in parallel. Motion is accomplished by shrinking and expanding the neighbors in the structure.

The current Projects for Self-reconfiguring Crystalline Robots include

  1. Creating distributed systems of Crystalline Robots
  2. Developing distributed algorithms for shape metamorphosis planning
  3. Developing algorithms that use self-reconfiguration for planning vestatile locomotion gaits; issues include mapping the terrain type to gait, keeping the static and dynamic stability of the shape, etc.
  4. Developing algorithms that use self-reconfiguration for self-repair
  5. Creating active sturfaces with Self-reconfiguring Molecule Robots

Generic Distributed Self-Reconfiguration Algorithms (cellular automata approach)

Our goal is to build robotic systems that are extremely versatile in both their structure and the tasks they can perform. This type of versatility is the hallmark of self-reconfiguring robots. Self-reconfiguring robots are systems composed of a large number of identical modules that can autonomously reshape themselves to fit the task at hand. These robots are massively parallel distributed systems, with each module thinking for itself within the group context.

Previous work in self-reconfiguring robots has concentrated on designing, engineering, and controlling particular systems. Several interesting robot designs have been proposed by Fukuda et. al., Murata et. al., Chirikjian et. al., Yim et. al., Shen et. al, Unsal et. al. and Hosokawa et. al., in addition to those developed at our laboratory. For each of these systems, architecture-dependent algorithms that couple planning to the specific actuation have been proposed. This includes real breakthroughs in both centralized planning and decentralized planning. This body of work has brought valuable insight into planning and control by focusing on designing and building hardware and developing algorithms coupled to specific hardware.

We believe we are now at a point where we can step back and examine more general questions about self-reconfiguration planning in an architecture-independent way. It is important to examine architecture-independent algorithms that can be instantiated for many different systems because they have the potential of providing a more general science-base for the field. By outlining general principles for reconfiguration planning, we hope to learn how to better design hardware and control algorithms. For example, the ability of a self-reconfiguring system to change shape can lead to "water-flow"-like locomotion algorithms that allow the robot to conform to the terrain on which it has to travel. Such algorithms have the potential of working
well in unstructured environments.

In most existing self-reconfiguring robot systems, an individual module can move in general ways relative to a structure of modules by traveling on the surface of the structure. Specifically, an individual module is capable of: (1) linear motion on plane of modules; (2) convex transitions into a different plane; and (3) concave transitions into a different plane. The details for how to accomplish these motions are architecture-dependent.

In our cellular automata-based work we introduce a new class of distributed control algorithms based on the above motion abstractions. These algorithms are generic, in that they can be instantiated for any self-reconfiguring robot where individual modules can move linearly on a structure of identical modules and make convex and concave transitions onto different planes of motion. Our control algorithms use local rules and are inspired by the cellular automata paradigm. Each locomotion algorithm employs a small set of local rules we implement as a cellular automaton. Each rule requires a set of preconditions on the neighborhood of the cell and when activated, causes the cell to move to an adjacent configuration. We represent the basic module of the robot as a cube, but our proposed abstraction can be replaced by any geometric structure that supports the formation of lattices.  

We illustrate our general approach to distributed control by developing algorithms for the locomotion task. Our abstract model permits us to prove that these algorithms produce reliable and correct locomotion, and allows the shape of the robot to comply to the geometry of the terrain in the spirit of the "water-flow" metaphor. We are also able to show that our correct abstract algorithms can be instantiated to very different robots with their correctness properties intact. For simulations of our cellular automata-based system go here.