In autonomous driving, bird’s-eye view (BEV) representations have emerged as the dominant approach for 3-D object detection. However, projecting 3-D objects into BEV space can lead to distant and nearby objects appearing similar in size, making it challenging to discern depth relationships between foreground and background objects. Furthermore, inadequate modeling of intermodal discrepancies and correlations hampers effective contextual integration in cross-modal fusion. To address these limitations, we propose hierarchical dynamic fusion (HD-Fusion), a novel end-to-end multimodal fusion framework consisting of a scene-level fusion (SLF) module and a contextual-level fusion (CLF) module. The SLF module fuses depth details from point cloud pillars with image features, generating BEV image representations enhanced with depth cues. The CLF module further enhances the features of LiDAR and cameras with a bidirectional cross-modal attention (BCMA) block and a discrete wavelet transform (DWT) encoder. The BCMA captures long-range interactions between LiDAR and image tokens, while the DWT separates multiscale frequency components to suppress noise and artifacts. Extensive experiments on the nuScenes benchmark show that HD-Fusion achieves 70.5% mAP and 72.9% NDS, improving over the baseline by 12.1 and 6.6 points, respectively. Additional evaluations on rainy/night subsets, simulated camera/LiDAR failures, and cross-dataset transfer from nuScenes to Lyft further demonstrate that HD-Fusion maintains superior performance on small and distant objects and exhibits strong robustness and generalization in challenging autonomous-driving scenarios.

Accurate attitude estimation is critical for navigation systems of uncrewed surface vehicle (USV) swarms, particularly under the challenges posed by cooperative operations in dynamic marine environments. In such swarms, significant misalignment errors arise not only from individual vessel motions induced by wind and waves but also from inter vessel kinematic coupling and asynchronous observation constraints, severely degrading cooperative navigation performance. To address this issue, this article proposes an attitude estimation method for a cooperative navigation system. By fully leveraging the advantages of the swarm, cooperative measurements are introduced as auxiliary information. Building upon the iterative cubature Kalman filter based on maximum a posteriori estimation, a novel measurement quality assessment approach utilizing a smooth boundary layer in the temporal-spatial dimension is designed to mitigate model mismatches in nonlinear conditions. A sliding window technique is employed to control the temporal span dynamically, adaptively adjusting the iteration count and computing an online correction for contaminated auxiliary information through a full-dimensional weighting matrix. Simulations and experimental results demonstrate that the proposed method effectively addresses model mismatches and achieves superior attitude estimation accuracy compared to existing approaches.

Cell-free massive multiple-input multiple-output (MIMO) systems have emerged as a transformative architecture for sixth generation (6G) communication networks, where distributed access points (APs) collaborate to simultaneously serve all user equipments (UEs). However, in time division duplex (TDD) systems, the reciprocity of uplink channel and downlink channel is disrupted by hardware imperfections in radio frequency (RF) chains, leading to significant degradation in system performance. This paper begins with a theoretical analysis of the downlink performance under a conjugate beamforming scheme, considering scenarios with and without channel calibration. A key theoretical insight highlights the limitation of conventional least squares (LS) calibration method, which fails to achieve high calibration accuracy even with an unlimited number of pilot observations. To overcome this limitation, we propose a novel channel calibration approach based on a diffusion model, designed to successively refine the calibration vector obtained from the LS calibration method. Furthermore, to address the shortcomings of conventional denoising diffusion probabilistic model (DDPM) training architectures, we introduce an innovative bridge-based diffusion model that maps the distribution of LS calibration vectors to their perfect counterparts. The proposed diffusion neural network architecture employs a conditional generative process, integrating a message passing neural network (MPNN) to incorporate domain-specific calibration insights. Numerical results demonstrate the superior performance of our proposed calibration method compared to existing methods, with supplementary experiments and in-depth analyses confirming the efficacy of the proposed successive refinement design.

Navigation systems are commonly used in Internet of Things (IoT) devices to provide navigation information. For achieving accurate information on devices, this article describes a robust variational Bayesian filter with sequential processing to estimate the time-varying measurement noise covariance and reject the outliers within the GNSS-challenged environment. Different from the variational Bayesian adaptive filter with heavy-tailed distribution, the proposed filter uses the beta-Bernoulli distribution to prevent the outliers from interfering with the time-varying measurement noise covariance estimation by the inverse-Gamma distribution. Then, the sequential processing is modified for the proposed filter in the independent multi-measurement sensor system. Sequential processing variational Bayesian interference is first proposed and applied to the tightly coupled navigation system, so far as the authors know. To evaluate the proposed method, real-time experiments in a downtown area and forest park effectively validated the filter robustness estimation, especially in scenarios with GNSS satellite obstructions.

Collision avoidance is critical for ensuring the safe navigation of unmanned surface vehicles (USVs). This paper presents an end-to-end solution for local path planning of USVs, focusing on enhanced obstacle evasion and smoother navigation. By leveraging deep reinforcement learning (DRL), we enable direct translation of relative distance states into navigational actions, eliminating the need for cumbersome map maintenance and complex feature extraction. A novel observation modality, the “beam map”, is designed to accurately perceive obstacles in all directions, mimicking the functionality of an onboard LiDAR system. To further refine collision avoidance maneuver, a warning zone is introduced, adjusting the agent’s sensitivity to obstacles and allowing ample time and space for decision-making. Additionally, we propose a continuous-time short-distance constraint to calculate the International Regulations for Preventing Collision at Sea (COLREGs) adherence rewards, enabling legal and rational navigation without requiring prior knowledge of the encounter situation. Extensive experimental results, comparing various RL policies and classical methods, demonstrate the planner’s exceptional obstacle avoidance capability and adaptability to changing environments. Using real-world inland ship navigation data, four steering scenarios are designed to further validate the efficacy of the proposed method.

Global navigation satellite systems (GNSS) often suffer from service interruptions or multipath errors in urban canyon environments, giving rise to reduced navigation accuracy. Therefore, it is necessary to develop effective fault-tolerant navigation systems to ensure a high-level accuracy despite GNSS failures. In this article, we present a novel fault detection framework based on the extended Kalman filter to address the problem of untimely fault detection and inaccurate positioning when GNSS fails. Specifically, we introduce the statistical process control technique of control charts to address the issue of slow-varying fault detection by constructing kernel multivariate exponentially weighted moving-average control charts instead of the conventional chi-square test. Simultaneously, we establish a corresponding criterion using EWMA-related statistics to mitigate the negative impact of uncertain noise and abnormal innovation, thereby ensuring the positioning accuracy of the navigation system. Finally, we validate the effectiveness and superiority of the proposed method through simulations and vehicle field data, demonstrating its ability to detect anomalies promptly and enhance the navigation and positioning accuracy while mitigating the adverse effects of GNSS lapse.

In indoor robot localization by using ultra-wideband (UWB), the extended Kalman filter (EKF)-based algorithms suffer from the colored measurement noise (CMN) that degrades the localization accuracy and causes the divergence. To overcome this issue, we develop a hybrid colored EKF and colored extended unbiased finite impulse response (EFIR) filter (cEKF/EFIR filter) employing measurement differences. We also develop this algorithm using a filter bank on merged averaging horizons to be adaptive to time-varying CMN and call it the adaptive EKF/EFIR (aEKF/EFIR) filter. Experimental testing is provided in UWB-based indoor mobile robot localization environments. It is shown that the end-to-end colored EKF/EFIR and aEKF/EFIR filtering algorithms have better performances than the EKF, EFIR filter, and their modifications for CMN.

The large misalignment angle errors generated by coarse alignment and the uncertainty time-varying calibration errors generated by the inertial measurement unit reduce the alignment accuracy, increase the alignment time, and ultimately limit the application of backtracking Kalman filters into In-motion fine alignment scene. This paper proposes a robust backtracking cubature Kalman filter (CKF) approach based on Krein space theory to overcome these issues. Specifically, considering the effect of dynamic model uncertainties of the alignment process, a linear robust filter and the existence conditions of optimal estimation are constructed to restrain the uncertainty interference according to Krein space theory. Meanwhile, an adaptive window adjustment algorithm is designed to intelligently determine the backtracking interval in different backtracking filtering stages and cross-scene motion environments, which is founded on the innovation variance gradient detection. Furthermore, using the statistical linearization scheme, the quasi-linear CKF model is derived to assist in resolving the nonlinear large misalignment angle within the Krein linear space framework for In-motion alignment. Experimental verification results from a vehicle In-motion alignment test illustrate that the proposed Krein backtracking CKF approach is effective in improving the alignment accuracy and shortening the alignment time simultaneously.