Clutaxis: an information-driven source search method balancing exploration and exploitation in turbulent environments

Abstract

Locating unknown emission sources in turbulent environments is a challenging yet crucial task, particularly in emergency response scenarios. Existing studies have developed information-theoretic approaches to fuse intermittent information collected by mobile sensors regarding the sources. This fused information is then used to support source-term estimation (STE) in various search algorithms. Among these, the cognitive strategy—a promising information-driven search algorithm—leverages a reward-based action selection mechanism to balance exploration and exploitation during each search step. However, this mechanism is hampered by a high computational load and rigid search trajectories, limiting its application in real-world systems. To address these issues, this paper proposes a novel information-driven search method called Clutaxis, based on a global exploration and exploitation tradeoff principle. Specifically, a particle filter is leveraged to maintain the STE. After projecting the particle filter samples onto a 2D search scene, the density-based spatial clustering of applications with noise (DBSCAN) algorithm is used to extract the density information of the samples, which is then used to construct a belief source area (BSA). By leveraging the uncertainty of the BSA, Clutaxis adopts explorative or exploitative actions with no restrictions on motion direction. Through dedicated simulations, the experimental results demonstrate the robustness of Clutaxis to key parameters and its advantages in computational complexity and search performance compared to two state-of-the-art algorithms (Infotaxis and Entrotaxis) and two Clutaxis variants (Clutaxis_ER and Clutaxis_EI).