A self-organizing, self-adaptive multi-agent system for the distribution of energy to a swarm of electric vehicles.

Abstract

In this thesis I construct a self-adaptive self-organizing multi-agent system that simulates the production and distribution of energy over a complex network to a swarm of electric vehicles. This network is comprised of autonomous agents which continually observe their immediate surroundings, reason internally and make decisions which have local effects. The main goal is to discover a heuristic for the behavior of distributor agents that enables distribution of energy from idle distributors to other distributors where demand exceeds supply with the intent of minimizing the Gini coefficient, a statistical measure of inequality between distributor agents, especially in simulation runs where total demand is lower than total supply, thus enabling very short charging times -- a term coined in the industry as “supercharging”, achieving output throttling, automatic load-balancing and, when demand exceeds supply, graceful degradation. Through a series of simulations I explore various scenarios to discover which heuristics work better than others in order to make reasonable data-driven decisions if we ever get to engineer in the future such a network in real life.