Skylark Cloud Corrections Cross-Country Drive: ©2020 Swift Navigation, Inc. All Rights Reserved
Skylark Cloud Corrections Cross-Country Drive: ©2020 Swift Navigation, Inc. All Rights Reserved
Skylark Cloud Corrections Cross-Country Drive: ©2020 Swift Navigation, Inc. All Rights Reserved
CROSS-COUNTRY DRIVE
Autonomous Navigation
Autonomous navigation continues to evolve and most car manufacturers have
now implemented programs involving various levels of autonomy. The key to
enabling full autonomous functionality for automotive applications lies within
the vehicle’s sensor suite. This suite consists of a combination (determined
by vehicle manufacturers) of radar, camera, LiDAR, inertial and GNSS (Global
Navigation Satellite Systems, of which GPS is a part). GNSS is a vital component
of the vehicle sensor suite as it is the only sensor to provide an absolute position,
velocity and time estimate. In addition, GNSS has a complementary performance
profile, providing the highest performance in environments in which other sensors
typically have difficulties or fail.
While various legacy GNSS positioning systems have been available for some
time, each have significant limitations and their broad adoption and application to
autonomous vehicles still require overcoming some significant challenges, such as:
Swift Navigation has created a precise positioning solution that addresses these
challenges. The Swift team undertook a drive test across the continental United
States (CONUS) to showcase the performance of Swift’s precise positioning
solution. Swift collected data over a range of geographies, using automotive-grade
hardware configurations, demonstrating the configurability and flexibility inherent
in the Swift solution. This drive covered 26 states and Washington D.C., with
6,614.7 miles (10,645.4 km) driven over 116 hours and 14 minutes logged by the
team at Swift.
This paper captures in detail the set-up and resulting data analysis that
demonstrates how Swift technology is the precise GNSS and sensor fusion leader
for automotive autonomy and that Swift’s positioning solutions meet the needs of
today’s—and tomorrow’s—autonomous vehicle manufacturers.
Vehicle Set-Up
A Swift test vehicle was equipped with 20 different GNSS devices utilizing five
unique chipsets. In addition, a number of different antennas were utilized,
ranging from a high-end, survey-grade antenna to a more price-conscious
automotive-grade antenna that would be representative of a typical production
Table 1: Devices
Receiver C Yes No
Automotive-grade antenna
Antenna-A
Represents current GNSS antenna on vehicles
Survey-grade antenna
Antenna-S
Represents higher-efficiency antenna
Route Driven
Many companies conduct drive tests in major metropolitan markets or along
heavily-populated coastal routes to demonstrate performance. Others drive
short or open-sky routes, utilizing ideal performance conditions. Swift chose to
undertake an ambitious cross-country drive test to showcase its technology both
in scale and performance. This report clearly demonstrates how Swift’s solutions
meet customer needs by operating over a large coverage area with consistent
and reliable precision. This drive test commenced on November 29, 2019 and was
completed on December 19, 2019 and traversed the continental United States,
covering 26 states and Washington D.C.
COVERED LOGGED
DROVE DRIVEN BY
100
80
40
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
ERROR (MET ERS)
3.000e+5
2.500e+5
Swift’s Piksi - Starling (DR) - Antenna-A
1.000e+5
5.000e+4
4
0.000e+0
0 0.5 1 1 .5 2
ERROR (MET ERS)
Key Takeaways
Variation of performance on different receivers with dual-frequency capabilities
• A 10% variation in drive time for 75 cm or less accuracy
• A 5.3% variation in drive time for 1 m or less accuracy
• Swift solution also achieved 1 m accuracy for a full 95.2% of the drive
• 75 cm of accuracy for 92% of the entire drive
These findings demonstrate that the Swift precise positioning solution can
achieve less than 75 cm of accuracy more than 92% of the drive time even without
additional standard vehicle inputs such as wheel odometry. With wheel odometry,
Swift has the capability to utilize sensor fusion to further improve accuracy.
Skylark Performance
100
80
PERCENT OF EPOCHS
60
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
70000
60000
50000
Swift’s Piksi - Starling - Skylark - Antenna-A
SA MPL E S
40000
Swift’s Piksi - Starling - SBAS Only - Antenna-A
30000
20000
10000
0 0.5 1 1.5 2
ERROR (MET ERS)
Key Takeaways
Swift’s Skylark performance improvements below with “Antenna-A” Comparing
Piksi Multi + Skylark with Piksi Multi + SBAS (without corrections) solution
80
P E RC E NT OF E POC HS
60
20
0 0.5 1 1 .5 2 2 .5 3 3 .5 4
1.000e+5
8.000e+4
6.000e+4
Swift’s Piksi - Starling - Skylark - Antenna-A
4.000e+4
2.000e+4
4
0.000e+0
0 0.5 1 1.5 2
ERROR (MET ERS)
Key Takeaways
Automotive-grade antenna impact on Swift performance compared to
survey-grade antenna
80
40
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
ERROR (MET ERS)
80
40
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
ERROR (MET ERS)
80
40
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
ERROR (MET ERS)
Atlanta to Dallas
100
80
40
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
ERROR (MET ERS)
80
20
0 0.5 1 1.5 2 2 .5 3 3 .5 4
ERROR (MET ERS)
Conclusion
This drive test took a Swift test vehicle through a wide variety of environments
as it traversed the United States. From snow in the Sierras of California, to the
streets of some of the world’s most bustling metropolises to open roads across
the continent. Under all conditions, Swift’s positioning solution demonstrated
its ability to consistently deliver the <75 centimeter accuracy required to keep a
vehicle safely positioned in its lane.
Since this drive test occurred, Swift’s engineering team has continued its work on
core positioning and sensor fusion algorithms to improve the availability, accuracy
and integrity of the overall Swift precise positioning solution.
For more information about how Swift’s precise positioning solution can benefit
your autonomous application contact Swift at sales@swiftnav.com.