Distributed Coverage Control for Spatial Processes Estimation with Noisy Observations

The present study addresses the challenge of effectively deploying a multi-robot team to optimally cover a domain with unknown density distribution. Specifically, we propose a distribute coverage-based control algorithm that enables a group of autonomous robots to simultaneously learn and estimate a spatial field over the domain. Additionally, we consider a scenario where the robots are deployed in a noisy environment or equipped with noisy sensors. To accomplish this, the control strategy utilizes Gaussian Process Regression (GPR) to construct a model of the monitored spatial process in the environment. Our strategy tackles the computational limits of Gaussian processes (GPs) when dealing with large data sets. The control algorithm filters the set of samples, limiting the GP training data to those that are relevant to improving the process estimate, avoiding excessive computational complexity and managing the noise in the observations. To evaluate the effectiveness of our proposed algorithm, we conducted several simulations and real platform experiments.