Deep Learning Based Decision Support Framework for Dead Reckoning in Emergency Vehicle Preemption

Dead reckoning, within the realm of emergency vehicle preemption, entails the art of deducing the present position and course of an emergency vehicle (EV), such as a police cruiser, ambulance, or fire engine. This deduction is rooted in prior knowledge and measurements, particularly vital when the accuracy of the Inertial Measurement Unit (IMU) is susceptible to decline amidst challenges like urban canyons, tunnels, or inclement weather. To surmount this challenge, we introduced a technique for estimating the EV’s position, termed “dead reckoning,” which leverages a deep neural network (DNN) in conjunction with an Inertial Measurement Unit (DNN-IMU). This self-contained system equips EVs with reliable navigation capabilities. In our initial phase, we designated six test routes, recording velocity, attitude (pitch and roll angle), and position data before integrating them with the DNN-IMU. These datasets underwent comprehensive training and testing. Throughout this process, we gauged the performance of the DNN-IMU using four key performance metrics, contrasting its effectiveness with prevailing methods. Simulation outcomes strongly suggest the efficacy of our proposed DNN-IMU across all six test routes. Notably, when tested in two different routes under GPS outage conditions, our method outperformed others, yielding significantly greater accuracy (92.45% for trajectory-1 and 93.62% for trajectory-2).