26734 research outputs found
Sort by
Lotterie pour batir des prisons, pour la Ville & District de Montreal
No full textExample of an early lottery ticket printed in eighteenth century Montreal
The compleat gamester : or, instructions how to play at billiards, trucks, bowls, and chess, 2nd ed.
No full textThe is thought to be the first English book devoted entirely to games and sports and was published anonymously and later ascribed to Charles Cotton
By the King, a proclamation for the better regulating of lotteries within the kingdoms of Great Britain and Ireland
No full textReal-Time UWB Indoor Localization: Deep Learning Approaches for NLoS Mitigation and Latency Compensation
No full textAccurate real-time indoor localization is essential for enabling safe, efficient, and intelligent operation in environments such as hospitals, factories, warehouses, and smart buildings. Applications ranging from autonomous robots to worker safety and asset tracking rely on precise and timely position estimates. Real-Time Locating Systems (RTLS) address this need by providing continuous spatial awareness through wireless technologies. Among RTLS options, Ultra-Wideband (UWB) stands out for its sub-meter ranging precision and robustness to interference. UWB receivers capture the signal propagation profile via the Channel Impulse Response (CIR), which reveals how signals arrive—both directly and through reflections from walls or obstacles. Under line-of-sight (LoS) conditions, the CIR’s first peak corresponds to the signal's true time-of-flight (ToF), enabling accurate range estimates. However, in non-line-of-sight (NLoS) scenarios—where obstacles block the direct path—reflected signals dominate, leading to overestimated distances and degraded positioning accuracy. This thesis introduces a deep learning framework to enhance both the accuracy and responsiveness of UWB-based localization. Neural models—including CNNs, BiLSTMs, and Transformers—are trained to learn the relationship between CIR patterns and range errors. These predicted errors are integrated into two geometric localization algorithms: Weighted Least Squares (WLS), where they adjust the relative importance of measurements, and the Extended Kalman Filter (EKF), where they dynamically scale the measurement noise. Both methods thus adaptively trust CIR measurements based on learned signal quality, improving robustness under multipath conditions. To address system latency, a delay-aware trajectory prediction framework is developed using deep sequence models—LSTM, Transformer, and CNN-Transformer hybrids—to forecast the agent’s future location from recent motion and CIR history. Predictions are aligned with actual system delay to ensure real-time accuracy. The resulting system, combining signal-level error correction with delay-aware forecasting, achieves robust UWB localization across diverse indoor environments. Experiments on a large-scale public dataset demonstrate consistent improvements in accuracy, generalization, and responsiveness, establishing deep learning as a practical foundation for precise, low-latency indoor positioning