ENGINES

The project considers the increasingly-relevant role played by artificial intelligence (AI) technologies in daily-life scenarios along with the overwhelming amount of AI techniques that engineers have at their disposal for Intelligent Systems Engineering (ISE). While engineers should have models and tools enabling to synergically exploit different AI techniques in an integrated way, this usually does not happen nowadays, as different AI techniques are mostly used either in isolation or with ad-hoc integrations tailored to a specific application—hence hindering both re-usability and conceptual integrity. Furthermore, whereas specific engineering methods and tools are available for specific techniques – such as for machine learning platforms or agent-oriented programming languages –, a comprehensive approach encompassing complementary AI techniques is currently missing. Agent-oriented models and technologies may help face this issue, as they offer abstractions, methods, and tools for ISE, by providing the common conceptual and technological grounding for their well-founded integration.

The overarching goal of the project is to define the ENGINES conceptual and technical framework providing models, guidelines, and tools for the integration of intelligent agent techniques, and in particular intelligent agent models and technologies with some selected AI techniques:

• coordination and agreement technologies
• conversation and natural language processing
• spatial modelling and reasoning
• semantic knowledge representation

chosen for their relevance in AI wrt currently-hot topics like explainability and understandability, as well as for the lack in terms of corresponding ready-to-use technologies. Here, other mature AI techniques – such as machine learning – could be used as off-the-shelf components in specific modules. In particular, we take as a reference the BDI intelligent agent logic and architecture as the de-facto standard for the engineering of intelligent agents capable of rational reasoning in dynamic environments. This makes it possible the engineering of intelligent systems based on existing and effective agent technologies augmented with the aforementioned AI techniques in a coherent, well-founded, comprehensive framework—instead of current practice based on ad-hoc solutions.

Accordingly, the main expected outcomes of the project are:

• an extended model and architecture for BDI agents augmented with the aforementioned AI techniques
• a set of guidelines for the integration of those AI techniques in the BDI architecture
• a toolchain for ISE based on the augmented BDI agents
• proofs of concept that demonstrate the feasibility and usability of the proposed model, guidelines, and toolchain.