The concept of autonomous vehicles dates back to the 1920s when the prototype of a radio-controlled self-driving car hit the streets of Manhattan, to the amazement of dumbstruck passersby. Ever since, the idea has captured the imagination of scientists and automotive engineers, marking the beginning of the ongoing race towards the self-driving future.
Today, the once distant vision of the fully autonomous car is inches away from reality. First driverless cars are roaming the streets in test condition, and ride-hailing giants are resuming their self-driving operations on public roads. And while we are yet to see the large-scale commercial deployment of Level 4 and Level 5 autonomous vehicles, highly automated and robotic systems have become commonplace in the luxury car segment, paving the way for fully autonomous capabilities.
The widespread adoption and price erosion of Advanced Driving Automation Systems (ADAS) as comprehensive driving assistance bring us significantly closer to the realization of self-driving cars. Nevertheless, ADAS alone isn’t sufficient to make autonomous cars a reality. To tackle one of the biggest challenges of autonomous driving — the ability to determine the exact position of a vehicle in real time — another critical enabler is required: high-definition mapping (HD mapping). Today, we’ll address the most common questions and challenges related to this technology.
High-definition maps: mapping the route to autonomous driving
What are HD maps?
As the action of driving gradually transfers from humans to machines, the role and scope of digital maps extends beyond navigation. GPS solutions cannot keep up with driverless cars, as they don’t provide data that’s dynamic and accurate enough for driverless vehicles. Autonomous driving software requires self-healing mapping systems explicitly built for self-driving cars to extend their vision and provide a detailed inventory of road features and objects on the side of the road.
The HD mapping technology allows a driverless vehicle to localize itself with high precision, mapping its exact location with respect to the surrounding environment. As opposed to ‘traditional’ maps intended for general navigation, HD maps for autonomous driving integrate and analyze sets of data from multiple sources, such as vehicle sensors, LiDAR, onboard cameras, satellite imagery, and GPS in real time. The fusion of this data reflects the exact position of the car in relation to all landmarks, supplying comprehensive, up-to-the-second information about road gradients and boundaries, traffic signaling, lane placement, anticipated curves, and safety conditions. As a result, high-definition maps deliver a faithful representation of the road with laser-sharp precision, setting a new standard for in-car navigation systems of autonomous vehicles. High definition maps and cloud data platform for autonomous driving
High definition maps and cloud data platform for autonomous driving
How does HD mapping impact the realization of self-driving cars?
High-precision maps increase reaction speed of AD systems already in place, which makes them a prerequisite for safe driving automation. By providing a 360-degree view of the road and driving conditions, HD maps allow self-driving vehicles to make instant decisions about driving strategy much faster than a human driver could. They improve sensor perception in extreme weather conditions or at a very close range (down to 10 cm), and are able to recognize objects and events that might otherwise go unnoticed by intelligent onboard sensors.
However, the ultimate goal of intelligent driving is to mimic the capabilities of a human driver, not only by ensuring safety but also by adding to passenger comfort. HD maps also help achieve that. Apart from their ability to dynamically map route conditions, advanced mapping systems aid self-driving cars with smooth path planning, allowing them to anticipate the road ahead and delivering superior passenger experience.
The vital challenges of high-precision HD mapping
Although HD mapping technology has made great strides in facilitating car self-localization and its implications are far-reaching, it still has some pitfalls that need to be addressed.
Large pools of data
HD mapping relies on compiling information from LiDAR scans, mobile cameras, connected sensors, and GPS devices, and updates it with crowdsourced inputs from commercial fleet partners. The captured data is collected into immense volumes, which must be instantaneously processed to produce immediate output. This imposes a burden on computing power. The fastest and most reliable way to resolve that issue is through comprehensive, cloud-based navigation platforms, which integrate map collection, aggregation and maintenance features. Based in the cloud, these features provide enough storage and high-performance computing power capabilities to support the demanding autonomous driving infrastructures.
Issues in data mapping
Unfortunately, some HD map data challenges remain unresolved. Geospatial data often comes in commercial or proprietary formats, and automobile companies and OEMs may lack the resources and technologies to collect it in its native format without quality loss. On top of that, each country has its own set of policies governing collecting, safeguarding, and disseminating spatial information. Many governments are yet to work on advancing their geospatial data activities. Until then, the sharing of geospatial data remains hindered in many cases.
Real-time data streaming issues
Speed is a crucial consideration in the pursuit of safety-critical driving automation. To deliver accuracy beyond conventional GPS solutions, high-definition navigation must bridge the gap between dynamic road changes and the time they appear on the map. Over-the-air updates of car positioning and road conditions require high-speed bandwidth and support for high vehicle density. The existing V2X networking and connectivity capabilities don’t provide these. Transmission allocation and cache allocation improvements are being made to reduce delays in HD map delivery, but latency remains an issue. However, 5G technology is anticipated to provide the resolution, offering transmission speeds of 5Gbps and latency below 10ms.
The difficulty in compiling high-precision HD maps within a reasonable budget is a common problem for mapmakers and autonomous vehicle manufacturers. Factors related to the persistently high costs of 3D mapping solutions are varied, but machine power and labor are some of the main culprits. Automotive companies, OEMs, and mapmakers strive to reduce expenditures in a way that doesn’t compromise technology quality at the expense of end product affordability. To accelerate their automotive breakthrough for driverless vehicles and reduce outlays, these companies are partnering with software engineering companies and working toward standardization.
Lack of common standards
The lack of a single automotive-grade navigation base is one of the crucial obstacles to the full commercial readiness and safety of self-driving cars. Individual smart car manufacturers, as well as navigation technology companies such as TomTom, HERE Maps or Nvidia, are publishing their own maps in proprietary formats; however, the real value of HD mapping can only be realized through standardization. Without it, map integrity and reliability are much more difficult to achieve. In fact, the NDS (Navigation Data Standards) Association was founded in recognition of the need for consistent data specification for HD mapping. Pipeline for compiling HD automotive maps
Pipeline for compiling HD automotive maps
Enhancing navigation data interoperability through NDS
What are Navigation Data Standards?
The NDS Association aims to create a global standard for in-vehicle navigation, ADAS, and e-horizon safety systems. The organization unites automotive OEMs, map data providers, and navigation device/application providers in a joint effort to build a single, integrated location reference that would ensure the integrity and reliability of HD mapping. Together, these automotive leaders are working toward a definition of a streamlined, globally adopted standard specification for a common method of storing and updating HD map data in a format that ensures easy access and interoperability and allows for seamless updates.
What is the value of NDS maps for autonomous driving platforms?
NDS navigation leverages mapping databases, which cover a dedicated geographical region and are maintained by a single supplier. The data within every base is organized into building blocks, each representing a particular functional aspect of navigation and layered with interconnected references. Since NDS maps work globally, they can be easily adapted to comply with country-specific regulations.
As a collaboration-based resource, NDS helps reduce the time and cost of developing a high-precision navigation system. It pools data from a global network of members, which means faster and easier map compilation. By compiling NDS-compliant HD maps, all stakeholders can redress the compatibility issues of trying to translate and interpret data gathered in non-compatible formats from disparate sources. Through data compatibility, the standard promises to speed up the feedback loop for immediate data updates in the system. A common framework for HD mapping also prevents vehicles from sharing data indiscriminately and sourcing it from unverified resources. In this way, it increases data quality and consequently enhances driving safety and experience.
Joining forces to enable full driving autonomy
A widespread rollout of autonomous driving is tantalizingly close, and HD mapping is among key contributors to its realization. However, the technology still has some challenges to tackle. Seeking to resolve them, advanced navigation companies such as TomTom, Nvidia, and HERE Maps are coming together and forging strategic alliances with specialty IT vendors like Intellias. Through collaboration, their aim is to alleviate cost burdens and accelerate market entry for their high-precision mapping solutions.
The Intellias team has expertise in building HD maps and NDS-based navigation for autonomous driving. We’re delivering impressive autonomous driving projects and taking charge of the full life cycle of 3D HD maps, from eliciting source data to creating and publishing the maps. Contact us to discuss your project requirements with one of our automotive engineers.