Run-Time Norms Synthesis in Dynamic Environments with Changing Objectives

Abstract

Normative Multi-Agent Systems (NorMAS) can model real-world applications as multi-agent systems and facilitate the coordination of the social behaviour of various entities (agents) interacting in an environment using norms. Aligning such norms with the objectives of the agents is crucially important to ensure that applying the norms would not affect the achievement of their objectives. However, when the environment is dynamic, agents can face unseen situations and might need to change their objectives accordingly. Therefore, it becomes more challenging to understand the change, synthesise norms, and align them with such dynamic objectives. This paper introduces a Dynamic Objectives and Norms Synthesizer and Reasoner (DONSR) model to align objectives and norms using a utility-based approach. An ontology-based schema, forward reasoning, and backward reasoning are used to identify the change in the environment and synthesise new objectives. Case-based reasoning enables the dynamic changing and reasoning of previously created objectives and synthesising norms. DONSR is evaluated using multiple simulated traffic scenarios, including different unseen situations (emergency events).Results show that norms can be synthesised and maintained efficiently while the objectives are being created and changed. Further, DONSR showed its efficacy in handling unseen situations, creating new objectives, and aligning them with the created norms.

NEMAS: Norm Entrepreneurship in Multi-Agent Systems

“…Norm synthesis requires a governing mechanism or a special centralised agent to generate and distribute the norms. Some noteworthy studies along this strand are the approaches of Shoham and Tennenholtz [9], Morales et al [10][11][12][13], Alechina et al [14], and Riad and Golpayegani [15][16][17]. These works commonly model the coordination problem as a simultaneous matrix game, as illustrated in Figure 1 [10].…”