Emergent Behaviors of a Linear Cluster Robot System: Enclosing and Molding by CA Method

This paper discusses emergent abilities of a cluster robot system and the design methodology of an algorithm for a behavior of the cluster robot system. A cluster robot system is an aggregation of simple robot modules, each including a locomotive device and several range sensors, that are interconnected by passive joints. Each robot module performs one of four types of behaviors, approaching , avoiding , staying , and complying . A cluster robot system can be designed from robot modules of same or mixed types. A cluster robot system interacts dynamically with its environment and other robot modules. A cluster robot system configured in a specific order of same or mixed types robot modules exhibits a speicific behavior according to that configuration. That is, a configuration of robot modules is an algorithm for a behavior of a cluster robot system. We call a configuration of robot modules a “ Configuration Algorithm (CA)” and this design method by CA the “ CA Method .” We assign a cluster robot a set of tasks, self-formation, enclosing an object, and molding an object. We experimentally demonstrate that emerging task abilities in a cluster robot system depend on the CA. The cluster robot interacts, for example, with an initially linear yielding object, then forms a stack of the object by our introduced emergent behaviors, pushing, wiping, buckling, and rolling-up. In another example, a cluster robot gathers all discrete objects within a large area based on the body size of the cluster robot.