Multi-robot On-line Sampling Scheduler for Persistent Monitoring

Abstract

The employment of autonomous agents for persistent monitoring tasks has significantly increased in recent years. In this sense, the data collection process must take into account limited resources, such as time and energy, whilst acquiring a sufficient amount of data to generate accurate models of underlying phenomena. Many different schedulers in the literature act in an off-line manner, which means they define the sequence of visit and generate a set of paths before any observations are made. However, on-line approaches can adapt their behavior based on previously collected data, allowing to obtain more precise models. In this paper, we propose an on-line scheduler which evaluates the sampling rate of the signals being measured to assign different priorities to different Points-of-Interest (PoIs). Next, according to this priority, it is determined if a region must be visited more or less frequently to increase our knowledge of the phenomenon. Our methodology was evaluated through several experiments in a simulated environment.