N P Patnaik M 1 1 ( October - December ) Current motor sports environments can generate vast amounts of telemetry data from cameras inertial measurement units IMUs pressure sensors tire temperature sensors rideheight sensors and vehicle dynamics systems Traditional diagnostics systems can have processing lag which can result in much longer reaction time under critical frontside time considerations In this paper a low latency sensor fusion architecture for motorsport realtime diagnostics with edge intelligence and multimodal sensor signal fusion is proposed The proposed architecture includes synchronised sensing adaptive Kalman filtering deep feature extraction and attention based predictive analytics with a very low endtoend latency and significant diagnostic accuracy The architecture features an attentionaugmented TDN Temporal Convolutional Network trained on a 100 proprietary motorsport telemetry dataset which includes 48 race sessions and more than 27500 annotated fault instances in eight fault categories Experimental results show that the proposed system processes the images with 844 faster time than the traditional cloudbased solution with only 28 ms of latency time on the edge platform NVIDIA Jetson AGX OrinExperimental evaluations demonstrate that a 986 accurate diagnostic tool can be executed using only 28 ms of latency time on an NVIDIA Jetson AGX Orin edge platform which is an 844 reduction in latency time compared to traditional cloudbased approaches With this system the inference speed is 145 frames per second which makes it possible to detect anomalies and predict faults in realtime when it comes to Formula 1 and endurance racing The proposed architecture is shown to be superior in all dimensions of performance through comparison with the thresholdbased CNN fusion and Transformer baselines Copyright (c) World Journal of Pharmaceutical Seiences. All rights reserved 2026