Multi-agent architecture for information retrieval and intelligent monitoring by UAVs in known environments affected by catastrophes

摘要

The consequences of natural or man-made catastrophes can be devastating. To minimize its impact, it is crucial to carry out a rapid analysis of the affected environment in the moments after they occur, especially from the perspective of alert notification or crisis management. In this context, the use of UAVs, understood as the technological basis on which intelligent systems capable of providing support to rescue teams is built, has positively contributed to face this challenge. In this article the design of a multi-agent architecture which enables the deployment of systems made up of intelligent agents that can monitor environments affected by a catastrophe and provide support to human staff in the decision-making process is proposed. These environments, known in advance, are characterized through a set of points of interests that are critical from the point of view of aerial surveillance and monitoring. To conduct an intelligent information analysis, a formal model of normality analysis is employed, which makes possible the definition of surveillance components. These represent the knowledge bases of the agents responsible for monitoring environments. Likewise, the architecture envisages communication and cooperation mechanisms between the different agents, as the basis for fusing information to assess the overall level of risk of the monitored environment. A case study is presented in which the spread of toxic smoke in an industrial complex which has just suffered a hypothetical earthquake is monitored.