A Spatially Informed Gaussian Process UCB Method for Decentralized Coverage Control
Gennaro Guidone, Luca Monegaglia, Elia Raimondi, Han Wang, Mattia Bianchi, Florian Dörfler
TL;DR
The paper presents a decentralized coverage control approach where each robot locally models an unknown density field with a Gaussian Process and selects motions via a GP-UCB–inspired loss that balances expected coverage with an exploration term. Scalability is achieved through sparse GP inducing points and a consensus-based hyperparameter update, while inducing points are chosen greedily to maximize information gain. Empirical results show the method outperforms centralized GP-based and model-based baselines and remains computationally efficient due to update scheduling and local communications. This work enables scalable, information-theoretically guided, multi-robot coverage in unknown environments with practical performance comparable to or better than model-based benchmarks.
Abstract
We present a novel decentralized algorithm for coverage control in unknown spatial environments modeled by Gaussian Processes (GPs). To trade-off between exploration and exploitation, each agent autonomously determines its trajectory by minimizing a local cost function. Inspired by the GP-UCB (Upper Confidence Bound for GPs) acquisition function, the proposed cost combines the expected locational cost with a variance-based exploration term, guiding agents toward regions that are both high in predicted density and model uncertainty. Compared to previous work, our algorithm operates in a fully decentralized fashion, relying only on local observations and communication with neighboring agents. In particular, agents periodically update their inducing points using a greedy selection strategy, enabling scalable online GP updates. We demonstrate the effectiveness of our algorithm in simulation.