EPOS

Information

EPOS, the Economic Planning and Optimized Selections, is a fully decentralized networked system designed for participatory multi-objective optimization forming a public good and supporting sharing economies. It performs collective decision-making among agents that autonomously generate a set of options from which they make a choice. Each agent is a human actor, a piece of software or a hybrid system of both that locally generates in a self-determined way a set of plan that define how some resources are allocated. For example, a plan may define the energy demand of a residential appliance in a future horizon or the availability of bicycles in the bicycle stations of a city. A set of several plans per agent represents alternative options, equivalent or not for the agent. This flexibility provides a degree of freedom in the overall aggregate allocation of resources in the system that results in a combinatorial explosion of possible trajectories of system-wise solutions: different combinations of local selections can lead to different desirable or undesirable global outcomes. EPOS is capable of steering such highly complex systems of combinatorial complexity to desirable outcomes by structuring agent interactions dynamic self-organized tree topologies and performing bottom-up collective decision-making using fitness functions designed to solve particular problems. for instance, preventing blackouts in smart grids by load-shifting or load-adjustment.

Both EPOS and external page I-EPOS can be applied to several application domains without changes in their core functionality. EPOS is primarily designed as a decentralized combinatorial optimization mechanism. I-EPOS, the Iterative EPOS system, adds decentralized back-propagation learning capabilities that improve system performance and the discovery of more efficient collective outcomes in an evolutionary fashion. Both EPOS and I-EPOS can be applied to several application domains without changes in their core functionality. Given its decentralization, scalability, local autonomy and collective decision-making, it can promote participation, fairness and sustainability in the sharing economies and application domains of energy, transportation, voting, Smart Cities and others.

Related Publications

Ilias Gerostathopoulos, Evangelos Pournaras, TRAPPed in Traffic? A Self-Adaptive Framework for Decentralized Traffic Optimization, in the Proceedings of the 14th International Symposium on Software Engineering for Adaptive and Self-managing Systems-SEAMS-2019, Montreal, Canada, May 2019 © IEEE (Awarded badge: Artifacts Evaluated – Reusable)

Evangelos Pournaras, Srivatsan Yadhunathan and Ada Diaconescu, external page Holarchic Structures for Decentralized Deep Learning-A Performance Analysis, Cluster Computing, 2019:1-22, Springer

Evangelos Pournaras, Peter Pilgerstorfer and Thomas Asikis, external page Decentralized Collective Learning for Self-managed Sharing Economies, ACM Transactions of Autonomous and Adaptive Systems, Vol. 13, Nr. 2, pp. 10, 2018, ACM

Evangelos Pournaras, Mark Yao, Dirk Helbing, external page Self-regulating Supply-Demand Systems, Future Generation Computer Systems, Vol. 76, pp. 73-91, 2017 © Elsevier

Peter Pilgerstorfer and Evangelos Pournaras, external page Self-adaptive Learning in Decentralized Combinatorial Optimization-A Design Paradigm for Sharing Economies, in the Proceedings of the 12th International Symposium on Software Engineering for Adaptive and Self-managing Systems-SEAMS-2017, Buenos Aires, May 2017