Exception Handling in Sistemi Multiagente: Il Caso di Studio del Linguaggio SARL

High level of abstraction and autonomy are the main strong points of the agent oriented programming (AOP) paradigm. Agents are autonomous software entities capable of complex decision-making based on perceiving the environment they're living into, indeed multi agent systems (MAS) are increasingly acquiring interest in many scientific and industrial fields, from simulation and modeling to self-driving cars. This work is focused on the study of SARL, a multi-purpose agent-oriented language which relies on a strongly multi-threaded structure and furnishes a default event-based communication standard between involved agents. SARL's key features are presented in the first part, while in the second attention has been particularly addressed to the exception handling and recovery mechanisms offered by the language, since faults managing constitutes one of the key features to build robust software. This aspect has been explored at different levels, starting from directly analyzing agents' threads to better understand the internal structure of SARL, to attempting to realize how exceptions could be modelled at a more abstracted level. The latter tries to explore the concepts of autonomy and responsibility in terms of a SARL agent, pointing out their implications in possible fault handling mechanisms, and realizing a set of differences between AOP and 'traditional' exception handling in object oriented programming. A sketch of a possible practical solution to a simple problem is also presented, enlightening its strengths and weaknesses and making emerge the conclusion of this work. In this latter aspect, the possible need to establish an accountability mechanism between agents is detailed, as a potential mean to build more reliable MAS.