Agent Process Modelling - Adaptive Multi-Agent System for Market-Based Regulation with Norms and Incentives
Keywords: Multi Agent Systems Regulation Norms Incentives Agent Process Modelling (APM)
Regulation is a control process in which management actions try to enforce system’s responses in order to achieve desired goals. When participants’ autonomy increases, such as in societal environments, regulation becomes a challenging problem. In these cases, shaping individual and collective behaviours is a management issue that can not be accomplished with traditional control approaches. Market-based regulation emerges as a feasible solution using economic and social artifacts, such as norms and incentives in order to attain a pro-active participation. From one side, regulation awareness is necessary for regulators to make regulation mechanisms effective. From the other, sensitiveness to regulation would help customers to decide whether to follow regulation or not. Tariffs and contracts express market rules acting as constraint enforcers. Since contracts can be violated and actors have different behaviours, adaptive regulation strategies are relevant and necessary. Indeed, such regulation will be more effective whenever the entities demonstrate responsiveness, i.e., having reasoning processes that encompass behavioural changes. Many real-life simulation scenarios lack responsiveness in agent models, from unresponsive behaviours and new adaptive regulation mechanisms could trigger a new path to automatic negotiation and integrated control strategies. In this thesis we tackle different application domains and related issues where distribution and autonomy are required. A suitable approach for autonomous and distributed entities: Multi Agent Systems (MAS). Moreover, we study how regulation can be achieved seamlessly in both simulation and real environments considering a new approach for intra-agent decision processes by following Business Process Modelling (BPM). We call this new approach Agent Process Modelling (APM). We aim to create a MAS that combines these concepts to create software agents that implement realistic user reasoning models and interact with the environment on the behalf of their principals. Application scenarios include smart grids, telecommunications, smart mobility, among others. The feasibility of our approach is mainly verified on the smart grid domain due to its relevance and applicability. We expect to allow more flexible and adaptive regulation processes in the studied scenarios and contribute by making it easier to design, manage and observe norms and incentives in MAS applications.