Switched Filter-Based Stationary Source Seeking by Multiple Agents

Accurate localization of an unknown signal source is critical in applications such as search and rescue, environmental monitoring, and defense surveillance. This paper presents a robust hybrid estimation framework that combines the Extended Kalman Filter (EKF) and Particle Filter (PF) within a dynamic switching architecture to estimate the position of a stationary source in a 2D plane using measurements from mobile sensors. The EKF is initially employed for its low computational cost, but switches to the PF when EKF reliability degrades due to nonlinearity and violation of Gaussian noise assumptions, as indicated by the Normalized Innovation Squared test. The system switches back to the EKF based on the Effective Sample Size metric to balance computational efficiency and estimation accuracy, enabling adaptive performance under nonlinear and non-Gaussian noise conditions. Additionally, a composite motion gradient for sensor-equipped agents’ motion planning is formulated leveraging information-theoretic and estimation-theoretic frameworks, thus balancing exploration and exploitation to maximize measurement informativeness. The proposed method is validated through numerical simulations, demonstrating efficient localization of a stationary source by a team of multiple agents.