Obtaining real-time sub-meter accuracy using a low cost GNSS device

Autonomous vehicles require accurate position at all times in different environments at an affordable price. This accurate position can only be achieved when combining multiple positioning methods. One of these methods is presented in this paper: positioning based on a Global Navigation Satellite System (GNSS) to obtain absolute position. This solution should be at an affordable price with sub-meter position accuracy. At the University of Delft, the Netherlands, a low cost solution was developed in Matlab for open areas which is called Single Frequency Precise Point Positioning (SF-PPP). It uses a low cost receiver with single frequency, single antenna and single GNSS constellation (GPS). The receiver provides raw measurements to the SF-PPP algorithm which corrects them for different kind of errors. This method was ported to a low cost Commercial Off-The-Shelf (COTS) embedded platform in C++. The selected platform is a Raspberry Pi version 2 with a u-Blox NEO 7P GPS receiver. The corrections for the raw measurements are received from a network service via a 4G modem. The PPP method is validated with an RTK system which is cm accurate. We evaluated the PPP method in different environments and conditions, with focus on open area, but also for harsh conditions on the highway and in an urban environment to know the current limitations of the method. For the open area environment a horizontal root mean square error (RMSe) of 0.5 m on position coordinates was achieved which fulfills our target of submeter accuracy. In harsh environments we suffer from reflections (caused by multipath receptions) and poor satellite availability due to obstructions from trees and buildings which makes the accuracy varying from 0.5 m up to 3 m. Future plans to improve the results involve using more satellites from other constellations like GLONASS, using the Doppler shift to estimate the vehicle speed, using dual frequency receiver for ionosphere removal and closer integration with other low-cost sensors and vehicle model.