Navigating through areas where GPS signals are unavailable, such as tunnels or underground parking structures, has long been a challenge for smartphone-based navigation systems. A collaborative team from Wuhan University and Chongqing University has introduced a novel solution to this problem. Their deep learning-enhanced framework, DMDVDR (Data- and Model-Driven Vehicle Dead Reckoning), enables smartphones to estimate a vehicle's position accurately without relying on GPS signals.
The system utilizes a custom-designed deep neural network, AVNet, to process data from a smartphone's inertial sensors and estimate the vehicle's orientation and velocity. This information is then integrated into an Invariant Extended Kalman Filter (InEKF) to compensate for sensor inaccuracies. Tested in real-world conditions, the framework demonstrated remarkable accuracy, with only 0.64% positional drift after 578 meters of GPS signal loss.
This advancement is significant as it provides a low-cost, scalable alternative to high-end vehicle navigation systems, which rely on expensive sensors. The DMDVDR framework has the potential to enhance various applications, including autonomous parking assistance and fleet management in GPS-denied environments. By merging artificial intelligence with classical control theory, this research paves the way for more reliable and intelligent navigation solutions using everyday consumer devices.
