Minimalism Distribution Supermodularity

Abstract

We have designed and implemented multi-agent strategies for manipulation tasks by distributing mechanically-based sequential algorithms across several autonomous spatially-separated agents, such as mobile robots. Our experience using mobile robots for the manipulation of large objects (couches, boxes, file cabinets, etc.) leads us to recommend a minimalist architecture for multi-agent programming. In particular, our methodology has led us to derive asynchronous distributed strategies that require no direct communication between agents, and very sparse geometric and dynamic models of the objects our robots manipulate. We argue for a design principle called supermodularity, which is orthogonal both to the notion of modularity in cognitive AI and also to horizontal decomposition (the non-modularity advocated in the subsumption /connectionist literature.) Finally, we discuss a simple mobot-Scheme infrastructure to implement supermodular architectures. In the past few years we have programm...