Deep Learning Breakthrough Enables Smartphone Navigation Without GPS
TL;DR
Researchers from Wuhan and Chongqing Universities have developed a smartphone-based navigation system that outperforms existing solutions in GPS-denied environments, offering a competitive edge in autonomous and fleet management applications.
The DMDVDR framework combines a deep neural network, AVNet, with an Invariant Extended Kalman Filter to accurately estimate vehicle position in GPS-denied areas using only smartphone IMU data.
This innovative navigation technology enhances safety and efficiency in tunnels and underground parking, making daily commutes and urban navigation more reliable for everyone.
A breakthrough in AI-driven navigation allows smartphones to guide vehicles through tunnels without GPS, merging deep learning with classical control theory for real-world reliability.
Found this article helpful?
Share it with your network and spread the knowledge!
Navigating through areas where GPS signals are unavailable, such as tunnels or underground parking structures, has long been a challenge for smartphone users. A collaborative team from Wuhan University and Chongqing University has introduced a groundbreaking solution to this problem. Their research, published in Satellite Navigation in June 2025, presents a smartphone-only inertial navigation framework named DMDVDR (Data- and Model-Driven Vehicle Dead Reckoning). This innovative approach leverages a custom-designed deep neural network, AVNet, to process data from a smartphone's inertial sensors, enabling accurate vehicle position estimation without relying on GPS signals.
The DMDVDR framework integrates AVNet's motion cues into an Invariant Extended Kalman Filter (InEKF), which compensates for sensor inaccuracies by combining model-based and AI-inferred data. A notable feature of this system is its data-driven filter parameter adapter, which dynamically adjusts to various driving conditions, enhancing the system's accuracy and robustness. Tested in real-world scenarios, including a parking lot and tunnel environments, the framework demonstrated exceptional performance, with minimal positional drift even in complex maneuvers like reverse parking.
This development has far-reaching implications for the future of navigation technology. By enabling reliable navigation in GPS-denied environments using only smartphone sensors, the DMDVDR framework offers a scalable and cost-effective alternative to traditional in-vehicle navigation systems. Potential applications range from autonomous parking assistance to fleet management in covered facilities, promising to enhance safety and efficiency in urban and underground navigation. As the demand for seamless indoor-outdoor localization grows, solutions like DMDVDR are set to play a pivotal role in the evolution of smart mobility.
The study, supported by the National Key Research and Development Program of China and the NSFC, represents a significant leap forward in AI-driven mobility. By harnessing the power of deep learning and classical control theory, the researchers have opened new avenues for navigation technology, ensuring that smartphones can provide accurate and uninterrupted guidance, even in the most challenging environments.
Curated from 24-7 Press Release
