Patterns of Delegate MAS

Abstract

Delegate MAS has been proposed and investigated as an integrated coordination technique for so-called self-organizing coordination-and-control applications. Delegate MAS consist of three types of light weight, ant-like agents that assist domain agents in their coordination tasks - the types are exploration, intention and feasibility ants. The technique is especially suitable for distributed applications in large-scale, dynamic systems. Literature shows that, for various application domains, solution approaches based on self-organization have been proposed that have several similarities to delegate MAS, yet are not identical. In this paper, we specify three reusable solution patterns for coordination in distributed, large-scale, dynamic systems. To motivate the patterns, we first visit several solution approaches from various domains that bear resemblance with respect to coordination. We then identify common application characteristics as well as common technical challenges that underlie the approaches. We describe recurring solution techniques, and consolidate these in three solution patterns. The patterns are "smart messages", "delegate MAS", and "delegate ant MAS" which rely on stigmergic ant agents. Identifying the patterns fosters reuse of particularly useful coordination techniques, and can serve as a catalyst for new or altered approaches.