Configuration Recognition, Self-Reassembly, and Distributed Fault Tolerance Studies with CKBot, a Modular Robotic System

Michael Park
Ph.D. Candidate
Advisor: Professor Mark Yim
Mechanical Engineering and Applied Mechanics
University of Pennsylvania

Abstract

We present developments in automating control for CKBot, our modular robotic system.  In particular, we discuss an implementation of an isomorphic configuration recognition algorithm for an arbitrary cluster of modules.  Applying basic concepts from graph theory, we introduce a centralized approach that allows a configuration of modules to detect its shape.  If a structure corresponds to a known shape (e.g., a particular configuration in a library of configurations), an on-board controller determines the exact permutation to execute the overall coordinated motion of the system.  We will also discuss the modular self-reassembly after explosion experiments.  This section will include information about the system’s hardware, hierarchical control architecture, and software state machine algorithm implemented in these experiments.  Lastly, we present a distributed, fault-tolerant control scheme based on a collective decision of modules in the presence of a minority of faulty members.  In this approach, all modules receive an infrared signal from a broadcaster; in certain cases a few modules may not see the signal, or may have faulty IR detectors.  We demonstrate that when a sub-critical number of modules are erroneous, a correct decision can still be made.