Robomotion: Scalable and Physically Stable Locomotion for Self-Reconfigurable Modular Robotics∗

Self-reconfigurable modular robots have an intriguingly flexible design, composing a single robot with many small modules that can autonomously move to transform the robot’s shape and structure. Each module may only have limited processing and memory resources. This makes scaling to a large number of modules quite challenging. This paper introduces a novel distributed locomotion algorithm for lattice-style self-reconfigurable robots which uses constant memory per module with constant computation and constant communication for each attempted module movement. Our major contributions in this paper is motion planning for self-reconfigurable modular robotics that is (1) scalable, (2) speedy, and most importantly, (3) stable. Our algorithm guarantees physical stability in the presence of gravity. By utilizing some robot modules to create a static support structure, other modules are able to move freely through the interior of this structure with minimal path planning and without causing instabilities or losing connectivity. This approach also permits the robot’s locomotion speed to remain nearly constant even as the number of modules in the robot grows very large. Our algorithm makes use of a distributed approach through coordination within small groups of modules. These groups are dynamically formed whenever they are needed, and disassembled when no longer needed. Additionally, we have developed methods to overcome dropped messages between modules or delays in module computation or movement. Empirical results from our simulation are also presented to demonstrate the scalability and locomotion speed advantages of this approach. ∗ This paper is a revised version of the conference proceedings paper: Sam Slee and John H. Reif, Robomotion: Scalable, Physically Stable Locomotion for Self-Reconfigurable Robots, Workshop on the Algorithmic Foundations of Robotics (WAFR 2010), Singapore, Springer (Dec. 13-15 2010). † Google Research, Google Corporation, 1600 Amphitheatre Parkway, Redwood Shores, California 94043, USA, e-mail: [email protected] ‡ Department of Computer Science, Duke University, Durham, NC 27707, USA, e-mail: [email protected]