Abstract
One of today's fast-growing domains in software engineering is multi-agent systems(MAS).One of the forces in MAS is the decentralised approach of achieving common, often complex, goals utilising simple agents.
At the same time, this is also a drawback because a lot of negotiation is required when agents engage in collaboration to achieve their common (complex )goals.
In this thesis the concepts; agent, activity and context, are investigated. And, it is investigated how to achieve collaboration between multiple agents by using activities in contexts,
while preserving a high level of agent autonomy. An experimental platform, a soccer simulator, labelled SoccerLab has been developed, to be able to experiment with agents and especially the concepts
activity and context and how they are applied during analysis, design and implementation.
The experiments revealed that activities are flexible and powerful for describing and prescribing collaboration, both at a conceptual level during analysis and design and
at a programming level during implementation, between agents in contexts, while preserving agent autonomy.