Abstract
Multi-agent frameworks are broadly available in the meanwhile. Nearly all of them however understand the agent concept differently, going as far as to provide several agent concepts within a platform. Along with this performance results of selected platforms allow to identify two clusters defined either by efficiency of processing or expressiveness of interaction language. As the only platform being related to both clusters the Otago Agent Platform (OPAL) which uses the efficiency-driven concept of ’Micro-agents’ is reviewed. As a result of the analysis the current implementation of the Micro-agents is considered to be too intransparent to allow monitoring at runtime or even cross-platform interaction. As such the implementation of a unified message infrastructure is suggested to simplify development of micro-agent interaction and provide monitoring capabilities while minimizing the loss of efficiency. The integration of Clojure as functional language is further considered to provide improved facilities for concurrent processing, usability and additional agent implementation languages.
The results of the hybrid imperative-functional implementation are behind the expected outcome as the runtime takes about five times as long as the original runtime. The results still offer potential for adoption as of the usability improvements and the fact that they are still far below the runtime of the slowest platform in the comparison. The choice of Clojure as language base for the whole platform is believed to be a promising next step to improve performance and avoid switching between different programming paradigms while keeping the improvements achieved. Further research with regards to a stronger integration of concepts of Agent-Oriented Software Engineering in the amended OPAL is suggested.