In multi-objective multi-agent systems (MOMASs), agents explicitly consider the possible trade-offs between conflicting objective functions. We argue that compromises between competing objectives in MOMAS should be analyzed on the basis of the utility that these compromises have for the users of a system, where an agent’s utility function maps their payoff vectors to scalar utility values. This utility-based approach naturally leads to two different optimization criteria for agents in a MOMAS: expected scalarized returns (ESRs) and scalarized expected returns (SERs). In this article, we explore the differences between these two criteria using the framework of multi-objective normal-form games (MONFGs). We demonstrate that the choice of optimization criterion (ESR or SER) can radically alter the set of equilibria in a MONFG when nonlinear utility functions are used.

The majority of multi-agent system (MAS) implementations aim to optimise agents' policies with respect to a single objective, despite the fact that many real-world problem domains are inherently multi-objective in nature. Multiobjective multi-agent systems (MOMAS) explicitly consider the possible trade-offs between conflicting objective functions. We argue that, in MOMAS, such compromises should be analysed on the basis of the utility that these compromises have for the users of a system. As is standard in multi-objective optimisation, we model the user utility using utility functions that map value or return vectors to scalar values. This approach naturally leads to two different optimisation criteria: expected scalarised returns (ESR) and scalarised expected returns (SER). We develop a new taxonomy which classifies multi-objective multi-agent decision making settings, on the basis of the reward structures, and which and how utility functions are applied. This allows us to offer a structured view of the field, to clearly delineate the current state-of-the-art in multi-objective multi-agent decision making approaches and to identify promising directions for future research. Starting from the execu-

Real-world sequential decision-making tasks are generally complex, requiring trade-offs between multiple, often conflicting, objectives. Despite this, the majority of research in reinforcement learning and decision-theoretic planning either assumes only a single objective, or that multiple objectives can be adequately handled via a simple linear combination. Such approaches may oversimplify the underlying problem and hence produce suboptimal results. This paper serves as a guide to the application of multi-objective methods to difficult problems, and is aimed at researchers who are already familiar with single-objective reinforcement learning and planning methods who wish to adopt a multi-objective perspective on their research, as well as practitioners who encounter multi-objective decision problems in practice. It identifies the factors that may influence the nature of the desired solution, and illustrates by example how these influence the design of multi-objective decision-making systems for complex problems.

Traditional support policies for green energy have greatly contributed to the rise in prosumer numbers.However, it is believed that they will soon start to exert negative impact on stakeholders and on the grid. Policy makers advise to phase out two of the most widely applied policies -net metering and feed-in tariff, in favor of support policies that scale better with rising renewable generation. This work quantifies the impact of these traditional policies in future "what-if" scenarios and confirms the need for their replacement. Based on simulations with real data, we compare net metering and feed-in tariff to four state-of-the-art marketbased mechanisms, which involve auction, negotiation and bitcoin-like currency. The paper examines the extent to which each of these mechanisms motivates not only green energy production but also its consumption. The properties and characteristics of the above mechanisms are evaluated from the perspective of key stakeholders in the low voltage grid -prosumers, consumers and energy providers. The outcome of this study sheds light on current and future issues that are relevant for policy makers in the evolving landscape of the smart grid.

Advances in reinforcement learning research have recently produced agents that are competent, or sometimes exceed human performance, in complex tasks. Most interesting real world problems however, are not restricted to one agent, but instead deal with multiple agents acting in the same environment and have proven to be challenging tasks to solve. In this work we present a study on a homogeneous open population of agents modelled as a multi-agent reinforcement learning (MARL) system. Using the SimuLane highway traffic simulator as a test-bed we show experimentally that using a single-agent learnt policy to initialise the multi-agent scenario, which we then fine-tune to the task, out-performs agents that learn in the multi-agent setting from scratch. Specifically we contribute an open population MARL configuration, how to transfer knowledge from single-to a multi-agent setting and a training procedure for a homogeneous open population of agents.