This paper presents an active observability optimization framework for mobile dynamic non-coincident ultra-wideband (UWB) localization systems. The method adopts a dual-pathway architecture that separates the real-time optimization of the tag’s geometric configuration from the recursive estimation of the platform’s kinematic state, thereby addressing the inherent orientation unobservability of conventional single-tag setups during degenerate motions such as straight-line travel. Specifically, a model predictive controller is first developed to actively reconfigure the tag’s radial distance and azimuth by maximizing a Fisher-information-based metric. Then, an extended Kalman filter incorporating the optimized geometry performs state estimation, where an observability-scaled gain formulation explicitly amplifies measurement updates for orientation when the configuration is favorable. For computationally constrained operation, geometric heuristics select the tag configuration based on the sensitivity of range measurements to orientation and the diversity of tag-to-anchor bearings. Simulations and real-world experiments demonstrate the effectiveness of the proposed framework in actively shaping the measurement geometry, while preserving real-time feasibility.
Unmanned Aerial Vehicles (UAVs) are increasingly used in intelligent transportation, disaster management, precision agriculture, military surveillance, and smart city applications because of their flexibility, mobility, and rapid deployment capabilities. However, resource-constrained UAVs often rely on edge-cloud collaborative infrastructures to execute computation-intensive tasks. While such architectures improve service availability and processing efficiency, they also introduce significant security challenges, including impersonation, replay, man-in-the-middle, and unauthorized access attacks. Furthermore, existing authentication schemes rarely address secure and resource-aware cloud server selection when edge resources become insufficient. To address these challenges, this paper proposes BAKE-UAV, a Blockchain-Assisted Authentication and Key Establishment scheme for UAV-enabled edge-cloud collaborative environments. In the proposed framework, a UAV first mutually authenticates with an Edge Server (ESE) and establishes a secure communication context. The ESE verifies the authenticity and integrity of the received request and evaluates its available computational resources. If sufficient resources are available, the ESE processes the request locally. Otherwise, the ESE submits a blockchain query using the service identifier and resource value to identify a Cloud Server (CSE) with adequate resources. The blockchain returns the selected server information and associated public parameters, enabling secure task delegation. The selected CSE retrieves encrypted service-related information from the blockchain, generates an authenticated response, and establishes a session key with the UAV. The proposed scheme integrates lightweight cryptographic primitives, including ASCON, ASCON-XOF, and Elliptic Curve Cryptography (ECC), to achieve strong security with low computational overhead. Security analysis demonstrates resistance against known attacks, while performance evaluation confirms that BAKE-UAV provides efficient communication and computation costs suitable for resource-constrained UAV deployments.
For sustainable smart tourism to thrive, efficient data-informed techniques must track, analyze, and anticipate the interplay of destinations, tourists, and environmental factors, yet real-world destinations require privacy when collaborating. We present a federated spatio-temporal digital twin for sustainable tourism forecasting: a virtual replica of a 24-destination panel coupled to observed physical states through a daily synchronization protocol, with PatchTST for temporal modeling, GraphSAGE for spatial dependencies, and Federated Learning with Differential Privacy (DP) for privacy-preserving distributed training. The twin operates closed-loop: destination-level states are refreshed from streamed observations, and a threshold-based feedback policy triggers operational interventions when projected sustainability deteriorates. On 24 worldwide destinations (2018–2024), our centralized benchmark achieves RMSE 4.81 ± 1.71 and MAE 4.14 ± 1.90, significantly outperforming STGCN (RMSE 8.16 ± 3.69) and Graph WaveNet (RMSE 6.38 ± 4.18) with a large effect size (Cohen’s d = 1.89, p < 0.05). Ablation shows explicit spatial encoding is necessary: a graph-less system yields RMSE 7.10, and air-route edges add value over geographic distance. Simulations with non-IID clients under DP with σ = 1.5 (ε = 11.3, δ = 10⁻⁵) achieve validation RMSE 6.58, outperforming strong standard regularization (6.79) and the non-IID baseline (6.99); relaxing to σ = 0.5 (ε = 59.4) further improves utility to 5.84 at the cost of a weak formal guarantee, with modest total communication overhead. In closed-loop operation, daily synchronization reduces cumulative 30-day state error from 6.41 to 2.38 RMSE relative to open-loop propagation, and the threshold-triggered replay reports an 8.54-point mean DSS difference whose intervention mechanism remains insufficiently specified for causal attribution. Spatial diagnostics report smaller between-destination differences in prediction error for connected pairs. This work provides a simulation-evaluated, privacy-preserving framework for multi-destination sustainable tourism forecasting with an operational virtual–physical control loop.
