MASAP: Multi-Agent Simulation of Air Pollution

摘要

This paper presents MASAP, a Multi-Agent Simulator of air pollution. The aim is to simulate the concentration of air pollutants emitted from sources (e.g. factories) and to assess the efficiency of air quality controlling policies. The pollutions sources are modelled as agents. Autonomously, the agents try to achieve their goals (increasing production, which has the side effect of increasing pollution) and cooperate with others agents by altering their emission rate according to the pollution level. The rewards/penalties are influenced by the pollutant concentration which is, in turn, determined using climatic parameters. In order to give predictions about the concentration of pollutants: Particulates Matter (PM10), Sulphur Oxide and Dioxide (SO_x), Nitrogen oxides (NO_x) and Ozone: (O_3), a combination between a GPD (Gaussian Plume Dispersion) algorithm and an ANN (Artificial Neural Network) is used. The prediction is calculated using real data about climatic parameters (wind speed, humidity, temperature and rainfall). Every agent cooperates with its neighbours that emit the same pollutant, and it learns how to adapt its strategy to earn reward and avoid penalties. When the pollution reaches a peak, agents are penalised according to their participation. The Simulator helps the decision makers to assess their air pollution controlling policies and oversee the results of their decision.