Coordinating mobile agents in interaction spaces

We present a kernel coordination language for mobile agent systems that considers as first-class citizens both the agents and the channels they use to interact with each other. Channels implement distributed, asynchronous communications with FIFO ordering and multicast routing. Features related to mobility include agent migration as well as remote cloning. Most importantly, a virtual form of channel mobility inspired by the image-calculus is also supported. This expressive feature allows mobile agents to adapt dynamically to their changing environment. The language semantics, presented formally, is based on a geometrical model named the Interaction Spaces. This provides an intuitive interpretation of the agent features and capabilities in terms of combined spatial projections and transformations. Through spatial composition, we show that standard labeled transition systems and bisimulation-based semantics may be defined above the geometry, enabling reasoning and formal verification. Finally, we describe prototype implementations of the proposed model and language.