GNSS/INS Sensor Fusion with On-Board Vehicle Sensors

A key requirement of autonomous vehicle applications is a reliable, accurate, and robust positioning (aka localization) solution. Key navigation, planning and decision operations cannot happen without dependable positioning. This means that accurate positioning must be ubiquitous - in other words, reliably available at all times and in all places the vehicle is expected to operate. While Global Navigation Satellite Systems (GNSS) commonly provide the basis for absolute positioning, it always suffers from the inherent problem of availability whenever a direct view of enough satellites is not possible. To address the failure mode, additional complementary sensors can be added to the overall navigation solution through a technique known as sensor fusion. Sensors such as inertial measurement units (IMUs), cameras, LiDARs, RADAR, etc. can be selected in such a way that the individual shortcomings of each sensor are mitigated, and the overall robustness and reliability are improved. Although current autonomous vehicle applications employ sensor fusion techniques, they tend to rely on high-performance sensors to meet the accuracy requirements. These high-performance sensors tend to induce a much higher cost burden than would be acceptable for commercial production, and therefore make mass autonomy too expensive. This paper will focus on the exploitation of the lower cost sensors already available on most modern vehicles. These sensors include low resolution odometry (DMI) and consumer grade IMUs currently used for dynamic stability control and wheel slip detection. A novel approach for combining vehicle speed, steering angles, transmission settings and multiple odometry inputs will be presented along with achievable results while operating under a GNSS denied environment. The test trajectory will mimic a typical parking structure with many corners and short straight segments. The only apriori information required for the filter is the wheel track and wheelbase (separation of wheels). A 90% performance improvement compared to the stand-alone GNSS/INS solution was observed during GNSS outages up to 30 minutes. Furthermore, up to a 50% improvement was observed when comparing between the multi-odometry vs single odometry outages during the same 30-minute outage condition. Beyond GNSS outage performance, it will be shown how the use of the extra input to the filter can improve protection levels of the positioning system to allow for more frequent engagement of the autonomous navigation system.