Source searching in unknown obstructed environments through source estimation, target determination, and path planning

Autonomous mobile robots have been gradually employed to search unknown sources in indoor environments. However, current studies have not fully addressed the source searching problem in unknown obstructed environments with limited sensing abilities. To deal with these problems, we propose an active source searching framework, in which mobile robots can avoid obstacles actively and realize the balance between exploration and exploitation in unknown obstructed environments through an iterative process: source estimation, target determination, and path planning. First, we describe the source searching problem and introduce the environment and sensor models. Then, a novel source searching algorithm based on particle filter, MEGI-taxis, and A-star is proposed. Specifically, the particle filter is used to estimate the source term parameters. The MEGI-taxis algorithm is developed to obtain a globally optimal searching target, which leverages the Gaussian Mixture Model to extract the features of probability information. Based on the heuristic rule, the A-star algorithm is employed to plan a collision-free path for the robot navigating to the target in unknown environments. When compared to state-of-the-art solutions in simulations, our method shows better performance in success rate, mean searching steps, and stability in the source searching process. Moreover, the effectiveness of the proposed framework is verified in the diffusion field generated by the computational fluid dynamics (CFD) model based on an indoor scene. The results reveal the important practicality of our proposed framework for source searching tasks in unknown obstructed environments.