Detecting and predicting environmental boundaries with a team of robots

Abstract

Large area disasters are usually triggered by small-scale anomalies in small areas which could possibly be detected in their early stages. The past decade has seen effective proposals to approach this problem with the deployment of mobile sensors for monitoring disaster prone areas. In this work, we study three important tasks to achieve an autonomous system that can monitor the environment and prevent catastrophes. The first task concerns the disaster detection. The challenge is to coordinate multiple robots to explore the environment in order to find anomalies. The second task concerns the subsequent disaster tracking. Once the anomaly is detected, the robots must coordinate themselves to track the behavior of the environmental boundary. In the third task, the resulting tracking information from the second task is used to estimate the current and to predict its future shape. The combination of these three tasks integrates a monitoring system that can alert and mitigate the risk suffered by human and animal beings. In this dissertation, we present some contributions for each one of these tasks and for their integration. We validate our proposed methods by simulations and with actual robots. Our experiments showed good performance results.