A real-time high-speed autonomous driving based on a low-cost RTK-GPS

In this paper, the feasibility of a low-cost real-time kinematic GPS (RTK-GPS) sensor for localization of an autonomous vehicle to achieve a high-speed driving is studied. For achieving high-speed autonomous driving, although the image-based method combined with a GPS can be a good approach for localization, the RTK-GPS position can be utilized with a low cost. On high-speed driving, it is important to acquire an accurate localization, because a less accurate position may degrade the performance to follow the waypoints on a target path with proper driving stability. Thus, in this study, the RTK-GPS position was applied to reduce position errors in a test vehicle with a low-cost GPS and the RTKLIB. A modified adaptive look-ahead distance pure pursuit algorithm was implemented to control the test vehicle. An autonomous driving experiment using the RTK-GPS position was carried out to verify the performance of the vehicle at a high speed of 130 kph on a track of 2477 m long with various corners and inclinations in a racing circuit. The test vehicle with the proposed real-time autonomous driving system using the RTK-GPS position achieved 111 s to complete a full lap on the racing track without departures from the track and noticeable lateral errors. This result was 32 s slower than the record accomplished by a professional human driver's 79 s.

Automated summary

This study examines the feasibility of using a low-cost real-time kinematic GPS sensor for localization of an autonomous vehicle to achieve high-speed driving. By utilizing the RTK-GPS position to reduce errors and implementing a modified pure pursuit algorithm, the test vehicle was able to complete a lap on a racing track at 130 kph without departures from the track or noticeable errors, albeit 32 seconds slower than the record set by a professional human driver.