Enhancing conventional search systems with multi-agent techniques: a case study

Our TECHS concept (TEams for Cooperative Heterogeneous Search) is aimed at the cooperation of different search systems. Therefore, a search agent is a sequential search system with a fixed setting of its parameters. In a team of agents, several agents might use the same search system but then they must employ different parameters. In contrast to most other approaches the problem to solve is not partitioned into subproblems that are then given to the search agents. Instead, all agents obtain the whole problem. Exceptions are only made for agents that cannot “understand” the whole problem. They only get the parts they can work on. Thus, we can use existing sequential search systems in our teams. The data interchange between search agents in TECHS is both demand driven and success driven. Demand driven cooperation is characterized by agents computing patterns of data that would help them fulfil their search task and send these patterns as requests to other agents. The other agents then send all their data matching the patterns (that may be the result of additional computations triggered by the request) back to the requesting agent. Note that in search systems it is rather difficult to find criteria that can determine which data an agent needs to advance or finish a search task. The general idea of success driven data interchange is that an agent determines the data that was, so far, very useful for its search and periodically shares this data with its colleagues