Co-ordination in the autonomous software agents' systems
J. Madejski · Journal of Achievements of Materials and Manufacturing Engineering · 2012
Purpose: Agents are designed to behave individually rational, which means that they should maximize their personal utility which is the way to make them less vulnerable to mean actions of others, yet they have to co-ordinate their actions to reach common goals, which is the purpose of this work. Design/methodology/approach: Agents can create and pursue their individual goals, behaving in a ‘selfish’ way to acquire the desired state of their world. To achieve that they may choose to adopt goals of other agents too, should this co-ordination be assessed as beneficial for them. Moreover, there is also a possibility to define the desired states of the agents in a way which will induce them to work together rather than try to operate individually. This may include their specialization, which forces in most cases sharing of their potential. This may be achieved by specialised design of agents being able to carry out elementary tasks. Such approach calls however, for design of a layer of supervisory agents which will be capable of realising what is the multi-agent overall system goal, setting up their teams from simple agents and committing to common sub-goals. All such systems may be efficiently developed only after careful study of the successfully operating systems in which humans are the agents, whose tasks may now be assigned to the software ones. These agents have to be coupled, as it also happens in their human counterparts. Findings: Development of the software agents’ co-operation framework based on review of publications covering both the fundamental considerations, as well as the latest developments. Research limitations/implications: Approach presented still needs careful testing and refinement of the co-ordination / negotiation rules.