28 February 2025 in Complex & Intelligent Systems
In this paper, we present a novel localization algorithm for mobile robots navigating in complex planar environments, a critical capability for various real-world applications such as autonomous driving, robotic assistance, and industrial automation. Although traditional methods such as particle filters and extended Kalman filters have been widely used, there is still room for assessing the capabilities of modern filtering techniques for this task. Building on a recent localization method that employs a chamfer distance-based observation model, derived from an implicit measurement equation, we explore its potential further by incorporating exact particle flow Daum–Huang filters to achieve superior accuracy. Recent advancements have spotlighted Daum–Huang filters as formidable contenders, outshining both the extended Kalman filters and traditional particle filters in various scenarios. We introduce two new Daum–Huang-based localization algorithms and assess their tracking performance through comprehensive simulations and real-world trials. Our algorithms are benchmarked against various methods, including the widely acclaimed Adaptive Monte–Carlo Localization algorithm. Overall, our algorithm demonstrates superior performance compared to the baseline models in simulations and exhibits competitive performance in the evaluated real-world application.
16 May 2026 in Knowledge-Based Systems
3 March 2026 in SSRN Electronic Journal
The Bayesian update step poses significant computational challenges in high-dimensional nonlinear estimation. While log-homotopy particle flow filters offer an alternative to stochastic sampling, existing formulations usually yield stiff differential equations. Conversely, existing deep learning approximations typically treat the update as a black-box task or rely on asymptotic relaxation, neglecting the exact geometric structure of the finite-horizon probability transport. In this work, we propose a physics-informed neural particle flow, which is an amortized inference framework. To construct the flow, we couple the log-homotopy trajectory of the prior to posterior density function with the continuity equation describing the density evolution. This derivation yields a governing partial differential equation (PDE), referred to as the master PDE. By embedding this PDE as a physical constraint into the loss function, we train a neural network to approximate the transport velocity field. This approach enables purely unsupervised training, eliminating the need for ground-truth posterior samples. We demonstrate that the neural parameterization acts as an implicit regularizer, mitigating the numerical stiffness inherent to analytic flows and reducing online computational complexity. Experimental validation on multimodal benchmarks and a challenging nonlinear scenario confirms better mode coverage and robustness compared to state-of-the-art baselines.