Abstract
This paper proposes an inertial sensor-based positioning method without using dead reckoning. The basic concept estimates the current position based on a pre-indexed map composed of spectrograms of road shapes estimated using inertial sensor signals. The proposed positioning algorithm has three characteristics: a feature-indexed map, the Kalman filter for signal fusion, and the interacting multiple model to integrate multiple maps. The positioning algorithm shows higher positioning accuracy and shorter computing time than a simple scanning method. Furthermore, the size of the map does not proportionally increase as the total length of the road increases. Therefore, the proposed positioning algorithm can be used as a complementary or alternative positioning method to GPS-based positioning methods using only an inertial sensor that is cheap and reliable.
Highlights
Vehicle self-positioning is essential for safe autonomous driving [1,2,3,4]
Some methods to improve the accuracy of Global Navigation Satellite System (GNSS) have been developed, including the Differential Global Positioning System (DGPS) [7] or Real-Time Kinematic (RTK) [8, 9], but they are too expensive to apply to production vehicles and even useless if GPS is completely blocked
This paper proposes an inertial-sensor-based non-dead reckoning positioning algorithm that can be applied in practice by realizing low computational load, consistent positioning accuracy, and high resolution and robustness
Summary
Vehicle self-positioning is essential for safe autonomous driving [1,2,3,4]. The most general positioning system is the Global Navigation Satellite System (GNSS), which provides information on the absolute position signals from at least four satellites, and users can access the position information with low-cost devices. The study presented how to estimate vertical road shape, demonstrated the repeatability of the vertical road shape at any vehicle type and speed ranges, and showed the possibility of positioning using the road shape spectrogram This method has many advantages, such as low cost, no error accumulation, and its use as a possible backup or alternative when other positioning methods are unavailable regardless of the vehicle types or speed ranges. This method still needs improvements for practical implementation, such as reducing computational load, increasing consistency of positioning accuracy, and simultaneously increasing resolution and robustness.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
