This research aims to enhance driver situational awareness during high-risk overtaking maneuvers by developing Navienta (Navigation Intelligent Assistant), a localized AI-powered navigation assistant. Conventional driver assistance systems often suffer from high latency and cloud dependencies that compromise real-time safety. To address these challenges, we implemented a localized edge-computing architecture utilizing a TF-350 LiDAR sensor and an Intel NUC as a processing hub, specifically designed to facilitate a high-speed, voice-driven interface. The system utilizes an Extended Kalman Filter (EKF) and a Mamdani Fuzzy Inference System (FIS) as the computational core to transform complex environmental dynamics into instantaneous voice instructions, ensuring low-latency feedback for the driver. The scientific contribution of this work lies in the synergistic integration of kinematic smoothing and fuzzy decision-making within a fully localized, high-concurrency architecture, eliminating cloud-dependency for safety-critical maneuvers. Experimental results confirm the system's precision with an average relative distance error of 0.22% and a consistent 50 ms end-to-end latency via a high-concurrency Golang backend. Experimental trials demonstrated that the localized Sherpa-ONNX engine achieved a 95.1% command recognition rate, which directly contributed to a significant 38.8% reduction in driver reaction time (from 1.8s to 1.1s). By maintaining operational integrity without external API dependencies, the Navienta framework provides a robust, cross-platform solution for modern intelligent transportation systems, offering a reliable approach for localized, safety-critical driver assistance.
Afiliasi berasal dari OpenAlex; metadata OAI-PMH jurnal tidak memuatnya. Penulis tanpa ID OpenAlex dicocokkan berdasarkan nama, sehingga dua orang bisa tergabung atau satu orang bisa terpisah.