GoalSPEC: A Goal Specification Language Supporting Adaptivity and Evolution

The characteristic of being autonomous and proactive makes the agents able to explore a wide solution space, that dynamically changes or contains uncertainty. We propose a language for describing system goals that may be injected at run-time into the system. The novelty of our approach consists in decoupling the business goals (what is expected) and their implementation (how to address the desired behavior). Indeed relieving the tension between ’what’ and ’how’ provides more degrees of freedom to the system. On the occurrence, agents of our system may exploit their features (mainly autonomy and proactivity, but also learning and planning) for getting benefits from a wider solution space. The result is that the system behavior may adapt to the current operating conditions. Moreover, the injection mechanism contributes to reduce the effort in evolving the system. This paper focuses on the goal specification language that is the base for enabling both adaptivity and evolution.