Swift Navigation is collaborating with Nvidia to enable a more scalable, cost-effective approach to autonomous driving by integrating the Nvidia Drive AGX platform with Swift’s globally referenced, centimeter-accurate GNSS positioning.
Swift Navigation offloads absolute localization to the GNSS sensor stack using its Swift Automotive Suite. The suite is a complete, modular software solution for safe, high-integrity precise vehicle localization that combines the centimeter-level Skylark Precise Positioning Service with the Starling positioning engine, software that fuses raw GNSS data and corrections with inertial sensors (IMU) and wheel odometry to deliver high-integrity, centimeter-accurate positioning (PVT).
By entrusting lane-level positioning to Swift’s high-precision stack, the vehicle’s optical sensors are relieved of the absolute positioning burden. This allows the perception stack to be optimized for obstacle detection and immediate safety, significantly reducing overall system cost and complexity.
Integration with Nvidia Drive AG
The integration is delivered through the Starling SAL Plugin for Nvidia DriveWorks. The Nvidia Drive AGX platform is the industry-standard, end-to-end platform for software-defined vehicles, scaling from assisted to fully autonomous operation. DriveWorks, its comprehensive SDK, provides a unified sensor abstraction layer (SAL) for seamless ingestion of data from all sensor types.
Swift’s new plugin acts as a drop-in component within this architecture. Sitting between the vehicle’s raw GNSS sensors and higher-layer software, such as that for localization, the plugin invisibly handles the complex mathematics of GNSS corrections and sensor fusion, outputting a clean, corrected position stream directly into the standard DriveWorks interface.
“We are removing the single biggest hurdle to widespread autonomy: the complexity and cost of localization,” said Holger Ippach, EVP of Product and Marketing at Swift Navigation. “By delivering Starling’s natively integrated, high-integrity GNSS to Nvidia DriveWorks, we are giving OEMs a direct path to globally referenced, lane-level positioning that is simple, scalable, and affordable.”
The collaboration and the Starling SAL Plugin unlock several advantages for automotive OEMs leveraging the Nvidia Drive platform:
Cloud-native ASIL safety. Skylark is an ASIL-certified positioning service built entirely in the cloud, offering scalability and reliability at a lower cost than solutions reliant on physical data centers.
Comprehensive sensor fusion. The Starling Positioning Engine delivers robust, high-integrity positioning by fusing precise GNSS with IMU and wheel odometry, ensuring continuous, lane-level accuracy even in signal-challenged environments.
Plug-and-play precision. Developers no longer need to build localization stacks from scratch. High precision is toggled on simply by adding the Starling plugin to the DriveWorks configuration.
Hardware independence. Because Starling is software-defined, Nvidia customers can achieve high performance using a wide variety of mass-market GNSS receivers, rather than being locked into expensive, proprietary navigation units.
Pre-validated integration. The Starling plugin has been rigorously tested and validated within the DriveWorks environment. This eliminates the complex, months-long burden of validating custom sensor drivers and fusion algorithms, allowing engineering teams to focus immediately on high-level path planning and control.
The Starling SAL Plugin for Nvidia DriveWorks is available now.
OxTS’ GNSS-aided inertial navigation systems (INS) are now supported on the NVIDIA DRIVE autonomous vehicle (AV) development platform. The software plug-in, developed in-house by OxTS using the NVIDIA DriveWorks SDK, runs on the NVIDIA DRIVE AGX Orin developer kit.
The plug-in gives developers using NVIDIA DRIVE the ability to feed OxTS GNSS/INS data directly into the platform to access accurate reference localization data as ground truth and validate the performance of the other sensors or algorithms under test.
A GNSS/INS is only one of several sensors required for an AV to operate. These sensors create a vast amount of data that must be synchronized, calibrated and centrally processed for the vehicle to operate safely. The OxTS GNSS/INS offers precision time protocol (PTP) time synchronization and can serve as the reference to calibrate all the other sensor data back for data analysis.
The OxTS RT3000 series is the GNSS/INS device supported on the NVIDIA DRIVE platform. It is currently being used as an advanced driver assistance system (ADAS) and vehicle dynamics ground-truth reference system for automotive test and validation teams across the globe.
The NVIDIA DRIVE platform is built on the DRIVE Orin system-on-a-chip and can process up to 254 trillion operations per second (TOPS) of sensor data from a variety of camera, lidar and radar devices.
Swiss company u-blox designs and manufactures GNSS receivers used in the automotive market, including driverless cars, and for micro-mobility devices, such as the Bird scooter.
In deep urban canyons, the biggest challenge for positioning cars is achieving sufficient accuracy despite multipath, said Aravinthan Athmanathan, product manager for the company’s Automotive GNSS line of receivers. “The challenge for autonomous driving is reliable lane-accurate positioning and integrity.”
The company develops its own dead-reckoning algorithms, which use data from an inertial measurement unit (IMU) and wheel speed sensors. “We also provide dual output, so the end customer can choose whether to use GNSS only or a sensor-fused solution,” said Athmanathan. This is especially challenging at the sub-meter accuracy level.
Different Uses, Different Sensors
Different automotive use cases require different GNSS receivers. To meet this challenge, u-blox offers the NEO-M9L for standard precision and the ZED-F9K for high precision, depending on the customer’s needs. Additionally, it is investing a lot “in functionally safe GNSS and in being the GNSS enabler for car manufacturers,” said Karin Steinhauser, the company’s senior marketing communications manager.
For navigation with meter-level accuracy, the NEO-M9L is integrated with dead-reckoning technology and sensor fusion, using algorithms that process sensor data from the IMU and from wheel-speed sensors. It can provide reliable location data in challenging environments, such as urban canyons, where multipath becomes an issue, or tunnels, where GNSS signals are partially or totally denied, Steinhauser said. Additionally, the NEO-M9L can operate in temperatures of up to 105° C, making it suitable for integration on the roof, behind the windscreen, or inside hot electronic control units. The NEO-M9L addresses the use cases in urban environments for both navigation and systems, such as Europe’s eCall, that provide an automated message to emergency services following a road crash, including the precise location of the accident.
The ZED-F9K, on the other hand, is well suited for use cases at the higher levels of advanced driver assist systems (ADAS) defined by the Society of Automotive Engineers (SAE), which require decimeter-level accuracy. “At L3 and above, you need correction services with integrity to allow for trustworthy and reliable GNSS positioning,” Steinhauser said. “We have partnerships with Bosch on projects to develop functionally safe GNSS solutions based on a ISO26262-certified version of u-blox generation 9 GNSS technology.” The ZED-F9K is a multi-band receiver that uses GPS signals on L1-L2 and Galileo signals on E5b. “We also have a special set of features adequate for the ADAS and the autonomous driving features,” Athmanathan said.
Image: 3alexd/E+/Getty Images
Bottlenecks
One of the factors limiting how quickly u-blox can roll out solutions based on the ISO 26262 standard (titled “Road vehicles – Functional safety”) is that highly autonomous systems require more integration work by the customers, said Alex Ngi, the company’s product manager for High Precision GNSS. “The first systems are now available.” Another hurdle, he pointed out, is the legal framework for deploying autonomous driving systems. “The regulations about how things need to be tested, and the liabilities for when systems fail, affect how quickly these systems can get adopted.”
GNSS can be used as a complementary technology to enable absolute positioning for systems that fuse data streams from cameras and lidars, such as those used for ADAS level 2 applications. “Fusing all this is computationally intensive and requires high processing power, such as NVIDIA GPUs, which tend to be very hot systems. We see a lot of requirements for very high-temperature GNSS receivers, because our receivers are often co-located with these hot systems.”
Of course, u-blox does not simply hand its modules to Bosch and car manufacturers and say, “You take it from here.” Design and integration is an iterative process. “We bring in the GNSS know-how and integration support and Bosch brings in the functional safe automotive development know-how,” Ngi said.
Dead Reckoning and Map Matching
For the automotive market, u-blox has more than 20 years of experience with dead reckoning. “The sensor-fusion solution receives data from both the GNSS and the IMU, and we provide the complete final solution,” Athmanathan explained.
The system also aids the receiver by providing it external map data. “If you’re driving your car northbound and the GNSS receiver tells you that it’s headed in the opposite direction, or that you’ve jumped over to the lane to the other side of the highway, clearly that cannot be right,” Ngi said. “Map matching relies on simple messages that come into our receivers to give us positive feedback on our measurements.”
For non-automotive applications, u-blox makes the ZED-F9R. It is used, for example, in robotic lawnmowers, very common in Asia and Europe, which require centimeter-level accuracies. “That’s why it focuses on delivering corrections using SPARTN, which can be a continent-wide data stream,” Ngi said. “We also make the design so that it’s very easy to integrate and enables the designers to easily pass the corrections to their receivers fully encrypted. This way, the value of the data is delivered to the lawnmower without exposing it to the system designer, so that we don’t need to go check every design to see whether somebody is leaking secured correction services.”
By the end of November, according to u-blox, updates of the ZED-F9P multi-band GNSS receiver will include decryption of the SPARTN correction data and a 95-percentile protection level. The protection level increases the trust non-safety-critical applications can place in its position output. By continuously outputting the upper bound of the maximum likely positioning error, referred to as the protection level, the receiver lets autonomous applications, such as UAVs or robotic lawnmowers, make efficient real time path planning, increasing the quality of their operations.
Guiding eScooters and EVs
In some places, Ngi pointed out, e-scooters are required to use a bike lane, which might be only two or three feet wide and may not be along the side of a building as it would be on a sidewalk. “The ZED-F9R is a much more flexible solution than camera systems that only know sidewalks or bike lanes.” Bird uses it to throttle driving speeds to match speed limits, which change from one location to another. “It is also much more scalable for them as opposed to such solutions as using UWB [ultra-wideband] beacons to fence off different areas, which are not really scalable for a company that wants to deploy solutions to hundreds of cities.”
Xpeng Motors, a manufacturer of smart electric vehicles, uses u-blox F9 GNSS receivers, which use signals from all four GNSS constellations, in its P7 super-long-range sports electric vehicle sedan. The vehicle uses ADAS for navigation-guided driving, automated parking and autonomous driving. For instance, once a navigation destination is set on a specific highway, the P7 will follow the route guidance to execute autonomous lane changing, switch to high-speed routes, and select the optimal route in real-time.
The PwrPak7-E1 from Hexagon | NovAtel is now supported on the Nvidia Drive Hyperion autonomous vehicle (AV) development platform. Selected for its robustness and precise position output, the PwrPak7-E1 will be offered with Nvidia’s autonomous driving test fleets worldwide.
Drive Hyperion is a fully operational, production-validated and open AV platform that reduces the time and cost required to outfit vehicles with autonomous driving and artificial intelligence (AI) features.
Powered by NovAtel’s OEM7 GNSS engine, the PwrPak7-E1 provides high-precision positioning used in the development of autonomous vehicles. The PwrPak7-E1 delivers NovAtel’s SPAN technology (GNSS + inertial navigation system, or INS) in an integrated, single enclosure.
Ground truth is the critical position reference for autonomous driving software behavior that can be validated. The PwrPak7-E1 provides ground truth in conjunction with Novatel’s Waypoint Inertial Explorer post-processing software. The device also has several connection options (serial, USB, CAN and Ethernet).
The GNSS and inertial measurement unit (IMU) output of the PwrPak7-E1, along with data from other onboard sensors, are recorded and fed into Nvidia’s sophisticated autonomous-driving development infrastructure and processing pipeline. There, data is synchronized, used for training AI models, and used in testing of various software components and autonomous driving behavior.
“Drive Hyperion is designed to give developers the ability to develop, evaluate, and validate AV technology more quickly,” explained Glenn Schuster, senior director of sensor ecosystems at Nvidia. “NovAtel’s compatibility on our platform provides developers the confidence to synchronize their sensor data with precision location information.”
Mobile World Congress (MWC) Barcelona will still take place Feb. 24-27, despite exhibitors dropping out because of the coronavirus. According to a statement by GSMA, some large exhibitors have decided not to come to the show this year because of the coronavirus, while others are still contemplating next steps. Despite this, the show is expected to boast more than 2,800 exhibitors.
According to Reuters, as of Feb. 9, there are two confirmed cases of coronavirus in Spain.
GSMA and its partners have implemented several safety measures for attendees, the organization said, including:
Increased cleaning and disinfection across all high-volume touch points, such as catering areas, surfaces, handrails, restrooms, entrances/exits and public touchscreens, along with the use of correct cleaning/sanitizing materials and products
Increased onsite medical support
Awareness campaign via online and onsite info-share and signage
Availability of sanitizing and disinfection materials for public use
Awareness and training to all staff on standard personal preventative measures, such as personal hygiene, frequent use of sanitizing/disinfection products, etc.
Advice to exhibitors on implementing effective cleaning and disinfection of stands and offices, along with guidance on personal hygiene measures and common preventive behavior
Public health guidelines and communication with Barcelona hotels, public and private transport, restaurants and catering outlets, retail, etc.
Installing new signage onsite reminding attendees of hygiene recommendations
Implementing a microphone disinfecting and change protocol for all speakers
Communicating advice to all attendees to adopt a “no-handshake policy”
A 24-hour telephone security and medical service for all attendees, available Feb. 12 to Feb. 29. This number will appear on the back of badge holders, in the event app and on signage around the venue.
Companies Back Out
Because of the coronavirus, several companies have backed out of attending MWC Barcelona. Several of these companies include Sony, Ericsson, LG, Nvidia and Amazon.
“Due to the outbreak and continued concerns about novel coronavirus, Amazon will withdraw from exhibiting and participating in Mobile World Congress 2020,” Amazon said in a statement. According to CNN Business, the company was due to host a dozen sessions covering topics such as 5G connectivity and artificial intelligence.
Other companies have expressed similar concerns.
“Sony has been closely monitoring the evolving situation following the novel coronavirus outbreak, which was declared a global emergency by the World Health Organization on January 30,” the company said in a statement. “As we place the utmost importance on the safety and wellbeing of our customers, partners, media and employees, we have taken the difficult decision to withdraw from exhibiting and participating at MWC 2020 in Barcelona, Spain.”
As an alternative, Sony’s press release will take place on Feb. 24 via a video through its official Xperia YouTube channel.
“We’ve informed GSMA, the organizers of MWC Barcelona, that we won’t be sending our employees to this year’s event,” Nvidia said in a statement. “Given public health risks around the coronavirus, ensuring the safety of our colleagues, partners and customers is our highest concern. We’ve been looking forward to sharing our work in AI, 5G and vRAN with the industry. We regret not attending, but believe this is the right decision.”
Rolls Royce is investing in autonomous shipping systems.
We no longer live in the Nuclear Age. Did you notice? Use of the term faded away. We no longer define our lives by the existence of nuclear technology, though of course it’s still part our world.
While the naming of “ages” is arbitrary, most people would say we’re currently in the Information Age. The last Great Age before the Information Age was the Industrial Age. We also experienced smaller “ages” such as the Space Age and the Atomic or Nuclear Age, but these didn’t impact our daily lives like the Industrial Age and Information Age have.
Which led me to wonder, what’s next? There’s a good chance we will find ourselves in the Autonomous Age, a daily experience of interacting with machines, robots and drones accomplishing tasks formerly done by people — including ourselves. We already use industrial robots; this trend will continue into new, more personal areas.
Inside our Smart Cities, we’ll wake up in a talking house built by autonomous construction machines, eat breakfast food that arrived via autonomous shipping and then delivered by drones, and travel to work via autonomous vehicles.
Artificial intelligence, augmented reality and autonomous vehicles will be woven into the tapestry of our daily lives. Overarching infrastructures and architectures will coordinate all the diverse autonomous and intelligent devices that we use.
“Artificial intelligence is sweeping across industries, and its next frontier is autonomous intelligent machines,” NVIDIA founder and CEO Jensen Huang said, speaking at GTC Japan in December. “Future machines will perceive their surroundings and be continuously alert, helping operators work more efficiently and safely.”
Komatsu plans to introduce NVIDIA graphics processing units (GPUs) to its SmartConstrution jobsites. The GPUs will communicate with drones from Skycatch, a Komatsu partner, which will collect 3D images, generate terrain data and “visualize” site conditions.
Komatsu is deploying the artifical intelligence (AI) project as an extension of its SmartConstruction initiative in Japan; the drone-assisted, automated equipment service was launched to alleviate the burden of the country’s severe shortage of skilled workers.
The company has deployed SmartConstruction at than 4,000 jobsites across the country, and the AI extension will be integrated into those sites.
Working with NVIDIA, OPTiM Corp. — another Komatsu partner and an internet of things management software company — will provide an application to correlate terrain data to jobsite workers and construction machines for visualization.
Enter Jetson. At the center of this collaboration is the NVIDIA Jetson artificial intelligence platform. When Jetson, which works with NVIDIA’s cloud technology, is installed in construction machines, it will be able to provide 360-degree images, enabling prompt recognition of workers and other machines nearby. The technology could potentially decrease fatalities that result from workers being struck by an object, piece of equipment or vehicle.
Jetson will also be used with the stereo cameras installed in the cabs of construction equipment, and will recognize continuously changing jobsite conditions on a real-time basis, to better provide accurate instructions to machine operators.
Future plans call for use not only for automatic control of devices, but also for high-resolution rendering and virtual simulation of construction and quarry jobsite operations.
Positioning and heading for mission-critical applications
The K528G dual-frequency, multi-constellation GNSS board provides the highest accuracy in differential positioning. It benefits from numerous constellation signals because of its advanced tracking performance of both GPS and GLONASS. The K528G can provide positioning and heading information generated by two antennas. It is designed for guiding and positioning construction engines, dredges, barges, shipping container cranes, mining equipment and intelligent transportation systems.
Designed for small-cell and distributed antenna systems
GPS Source has released of a line of GPS/GNSS splitters created for the small-cell wireless and distributed antenna system markets. Specifically designed for the L-band frequency, they can eliminate the cost of multiple antennas and long cable runs in wireless installations. With four or eight outputs, the new line of splitters make it possible to use a single GPS referencing antenna and cable arrangement for multiple synchronized systems. The splitters include features such as DC bias select and amplification. GPS Source RF signal splitters typically operate in conjunction with an active GPS antenna; consequently, a GPS RF signal splitter must have provisions for managing the DC voltage to the active GPS antenna. The S14GT and S18GT splitters will power an external GPS antenna from any of the RF outputs. A “hunt-and-pick” circuit is used to select only one DC input for power should more than one source be connected. Designed for redundancy, if the selected DC bias input should fail, the DC bias will automatically switch to another DC input to ensure an uninterrupted power supply to the active antenna.
For precision industrial, agricultural and military OEM applications
A new series of L1 band wideband antennas for OEM applications is offered in three formats:
▪ TW2106/TW2108 — GPS L1
▪ TW2406/TW2408 — GPS + GLONASS
▪ TW2706/TW2708 — Galileo, BeiDou, GPS + GLONASS
Each antenna type features Tallysman’s Accutenna technology, which provides high rejection of multipath signals, with low axial ratios and tight phase center variations (PCV). Each is available with a brickwall pre-filter option to protect against saturation by high level subharmonic and L-band signals. The antenna printed circuit boards (PCBs) are 56 millimeters in diameter with four plated holes for secure mounting. They are available with a variety of connectors and custom cable lengths, and can be custom-tuned. All of them are REACH and ROHS compliant.
The NV08C-RTK-A is fully integrated multi-constellation L1 heading receiver with embedded real-tiime kinematic (RTK) functionality and compatibility with GPS, GLONASS, Galileo and BeiDou. The NV08C-RTK-A is designed for use in high-accuracy applications that demand low-cost, low-power consumption, a small form factor and high performance, such as construction, mining and industrial; environmental and structural monitoring; machine control; parallel driving systems; precision agriculture; UAVs; and robotics and intelligent machines.
The SLD-100 GNSS Rover accessory facilitates hydrographic measurement in bodies of water up to 100 meters in depth. it is designed for anyone who finds themselves needing to survey into bodies of water, streams and rivers. With survey-grade accuracy, the SLD-100 can be added to any brand GNSS RTK rover to allow for position and depth measurements to be made simultaneously. With a built-in 10-hour lithium battery and transmitter unit with Bluetooth connectivity, the SLD-100 provides standard-depth data streams in several industry-standard NMEA formats at 1 Hz, 4800 bps, providing compatibility with any hydrographic surveying software package. Position and depth information is externally logged on a computer or controller. Included transom mounting hardware enables easy installation.
TriAnt is small, thin and rugged high-performance GNSS antenna. It measures 128 x 128 millimeters (mm) square and 39 mm thick. It can be mounted with three screws to flat surfaces. It is designed for applications such as machine control and surround anennas of the TRIUMPH-4X. The antenna cable is routed through the center of the antenna (TNC connector) for protection in harsh environments. The TriAnt can also be mounted on poles (1–14 inches thread) using its mount-pole attachment, which increases the thickness to 54.5 mm.
The X20i L1 GPS receiver by CHC Navigation is powered by a high-precision L1 GPS engine. Its integrated Bluetooth chip enables it to wirelessly collect submeter positions in real- time or centimeter post-processed on an iPhone or iPad. All location-aware apps on the iPhone and iPad are compatible with the X20i. Immediately after pairing and answering the security question allowing the X20i to take control of location services on the iOS device, 1 million iOS applications are capable of utilizing the high-accuracy data of the X20i, and become accurate to either 1 foot or 1 centimeter. Apps that can make use of the high accuracy include TerraGo Edge, ESRI’s ArcView Connector and those by CarteGraph Systems.
BlueStarGPS offers both GPS and GNSS options in a rugged, lightweight package. The BlueStarGPS device was designed to meet sub-meter mapping and data-collection needs in the pipeline and utility industries. It provides sub-meter precision without post-processing, and maintains accurate positioning when the SBAS signal is obstructed. This means it can function under trees, around buildings and in rugged terrain where other receivers can fail. The BlueStarGPS is designed specifically for use with Android mobile devices, such as smartphones, tablets or notebook computers, as well as cable and pipe “locating” tools with a connectivity range of up to 1 kilometer.
UAV measures through water surfaces of rivers, lakes
The RIEGL BathyCopter is a small-UAV-based surveying system capable of measuring through the water surface. It’s suitable for generating profiles of rivers or water reservoirs. The platform design integrates a topo-bathymetric green laser depth meter, an APX 15 inertial measurement unit (IMU)/GNSS with antenna, a control unit and a digital camera. Applications include generation of river profiles, survey of reservoirs and canals, landscaping, support of construction projects, and surveys for planning and carrying out hydraulic engineering work.
The Zenmuse X5 is a micro four-thirds (M4/3) camera designed specifically for aerial use. With a large sensor, aerial image makers will be able to capture up to 13 stops of dynamic range, enabling capture of high-resolution 16-megapixel photos or 4 k, 24 fps and 30 fps videos in complex lighting environments. It supports four interchangeable lenses. The Zenmuse X5 is designed for creation of high-quality aerial maps and 3D models, industrial and utility inspection, and professional video capture.
The NVIDIA Jetson TX1 module is designed to power smart devices — including drones that don’t just fly by remote control, but navigate their way through a forest for search and rescue. It is an embedded computer designed to learn to recognize objects or interpret information, incorporating capabilities such as machine learning, computer vision and navigation into a single system. This technology expands the ability of machines to operate on their own and adapt to their surroundings by recognizing images, processing conversational speech, or analyzing a room full of furniture and finding a path to navigate across it.
