Multi-Agent Clarity-Aware Dynamic Coverage with Gaussian Processes
Devansh R. Agrawal, Dimitra Panagou
TL;DR
The paper tackles scalable, informative multi-agent coverage of spatiotemporal fields by modeling the environment with a separable spatiotemporal Gaussian Process and tracking uncertainty via a clarity metric. It derives a Kalman-filter based information assimilation framework and two GP-informed coverage controllers, one optimizing direct clarity gain and the other leveraging ergodic control to meet a target clarity distribution. The approach is demonstrated through wind-field simulations over a real Austrian region, showing that multiple robots can efficiently increase and maintain domain clarity while producing accurate reconstructions. The work highlights a principled connection between GP models, SDEs, and ergodic control to enable scalable, centralized data fusion and distributed, clarity-driven exploration. Limitations include fixed GP hyperparameters and centralized fusion, with future work pointing to online hyperparameter learning and safety-aware, distributed implementations.
Abstract
This paper presents two algorithms for multi-agent dynamic coverage in spatiotemporal environments, where the coverage algorithms are informed by the method of data assimilation. In particular, we show that by explicitly modeling the environment using a Gaussian Process (GP) model, and considering the sensing capabilities and the dynamics of a team of robots, we can design an estimation algorithm and multi-agent coverage controller that explores and estimates the state of the spatiotemporal environment. The uncertainty of the estimate is quantified using clarity, an information-theoretic metric, where higher clarity corresponds to lower uncertainty. By exploiting the relationship between GPs and Stochastic Differential Equations (SDEs) we quantify the increase in clarity of the estimated state at any position due to a measurement taken from any other position. We use this relationship to design two new coverage controllers, both of which scale well with the number of agents exploring the domain, assuming the robots can share the map of the clarity over the spatial domain via communication. We demonstrate the algorithms through a realistic simulation of a team of robots collecting wind data over a region in Austria.