In multi-objective multi-agent systems (MOMAS), agents explicitlyconsider the possible tradeoffs between conflicting objective func-tions. We argue that compromises between competing objectivesin MOMAS should be analysed on the basis of the utility that thesecompromises have for the users of a system, where an agent’s util-ity function maps their payoff vectors to scalar utility values. Thisutility-based approach naturally leads to two different optimisationcriteria for agents in a MOMAS: expected scalarised returns (ESR)and scalarised expected returns (SER). In this paper, we explorethe differences between these two criteria using the framework ofmulti-objective normal form games (MONFGs). We demonstratethat the choice of optimisation criterion (ESR or SER) can radi-cally alter the set of equilibria in a MONFG when non-linear utilityfunctions are used.
|Title of host publication||Adaptive and Learning Agents Workshop 2019 (ALA-19)|
|Publisher||Vrije Universiteit Brussel|
|Number of pages||9|
|Publication status||Published - Jun 2019|
|Event||Adaptive and Learning Agents (ALA) Workshop 2019|
at AAMAS, Montreal, Canada - Montreal, Canada
Duration: 13 May 2019 → 14 May 2019
|Conference||Adaptive and Learning Agents (ALA) Workshop 2019|
at AAMAS, Montreal, Canada
|Period||13/05/19 → 14/05/19|
Rădulescu, R., Mannion, P., Roijers, D. M., & Nowé, A. (2019). Equilibria in Multi-Objective Games: a Utility-Based Perspective. In Adaptive and Learning Agents Workshop 2019 (ALA-19) (pp. 1-9). (AAMAS series; Vol. 11). Vrije Universiteit Brussel.