Benefits of Mapless Autonomous Driving Technology

by Ilan Shaviv, Ph.D. (CTO) | Imagry

When it comes to the development of autonomous driving technology, the role of HD (High-Definition) maps has been central to the discussion for years. Many companies have invested significant resources in creating and maintaining these HD maps, believing they are a crucial component of the self-driving puzzle. However, at Imagry we challenge the notion that HD maps are the right solution for enabling vehicles to drive autonomously. It is our belief that the future of autonomous driving is not paved with HD maps, but rather will be achieved through adaptable and dynamic AI architecture capable of navigating existing road infrastructure. In this article, we will review the flaws inherent in the HD maps approach and make the case for going HD-mapless.

What are the flaws inherent in the HD-mapping approach to autonomous driving?

📉 Expensive Creation and Maintenance
HD maps are expensive. Creating detailed, high-definition maps of an entire city or region is a time-consuming and resource-intensive process. Specialized mapping companies often spend years gathering and processing data to create these maps, and maintaining their accuracy becomes an ongoing challenge. The discrepancy between HD maps and real-world conditions mandates costly updates and constant maintenance to keep these maps relevant. These creation and maintenance costs, passed on to the OEM or Tier-1, are prohibitive.

🕐 Lack of Real-Time Information
A significant drawback of HD maps is that they do not represent real-time conditions on the road. Autonomous vehicles need to be aware of dynamic situations, such as detours, road closures, construction, various objects on the road, traffic accidents, and more. HD maps cannot provide this real-time information, leading to potential issues when autonomous vehicles encounter unexpected roadblocks or obstacles. The mismatch between HD maps and the current state of the road can lead to confusion and challenges in decision-making for the self-driving vehicle, creating new hazards (much to the frustration of human drivers who share the road).

🤝 Dependency on Vendors
Relying on HD maps can create a dependency on the vendors who produce and maintain them. There are only a few companies attempting to map the entire world, and their efforts have been met with limited success. This dependence on third-party entities can limit the flexibility and adaptability of autonomous driving systems.

💽 Data Storage Challenges
Storing HD maps locally on autonomous vehicles poses a variety of challenges. These maps can be incredibly large in size, making it impractical to download them over the air and store them on autonomous vehicles. For example, downloading a high-definition map of a large city like Los Angeles or New York would require several terabytes of storage, which is not economical for most vehicles.

📡 Communication Limitations
While cloud communication is a potential solution to avoid storing large HD maps locally, it is not always a reliable option. Continuous high bandwidth is required for seamless communication, and this can create technical challenges, especially in areas with limited bandwidth coverage, network congestion, or tunnels. It is also quite expensive to maintain, and it is not reasonable to expect that these ongoing charges would be borne by vehicle owners!

cyber vulnerability Cyber Vulnerability
An autonomous vehicle using HD maps for navigation gets its driving directions from the outside via an open communication port. Therefore, it is prone to cyber-attacks.

🚜 Irrelevance for Off-Road Applications
HD maps are primarily designed for regular public roads in urban, suburban, and highway environments. But these HD maps are completely irrelevant for off-road applications, such as mining or agriculture, where the terrain and conditions are vastly different. In these scenarios, the road infrastructure does not exist as it does in city settings, rendering HD maps useless.

📍 Localization and Perception Challenges
Using HD maps introduces the challenge of localization, where autonomous vehicles need to determine their exact position on the map. GPS signals in cities can be unreliable due to a variety of effects like “GPS canyon” and underground passage, to name a few, whereas using a compass for heading is not always accurate. Reconciling the differences between the HD map’s predictions and the actual environment requires a perception system, which can again lead to uncertainty and decision-making dilemmas (e.g., Which system to believe? The HD map or the perception layer?).

What are the Options? HD Maps vs. Real-Time Perception

HD maps are used to describe how the world looks to the motion planning module. In order to know which section of the map to use, a good localization system like GNSS augmented with RTK (Real Time Kinematics) corrections, is required. Information about dynamic objects (e.g., other vehicles, pedestrians, etc.) which are not included in the static HD map data, is also necessary.

The HD-Mapless Approach: Real-time Image Recognition

Our real-time image recognition system uses the video feed from multiple onboard cameras to produce a reliable top-down view of the surrounding environment. Visible Imaging Sensor (“VIS”) cameras constantly scan 360˚ for distances up to 300 meters from the vehicle. The system detects and perceives road geometry and markings, traffic signals and signs in real-time, and tracks various objects to predict their trajectory and velocity, delivering critical input of surroundings for the motion planning stage.

Bio-Inspired Technology for Location Independence

Much like a human driver, the Imagry-enabled vehicle is not limited to known (“mapped”) roads. Instead, it navigates unfamiliar areas based on real-time perception and memory of previous driving behaviors. That is why we call it “location-independent”.

Economical, Self-Sufficient Computing Platform

Furthermore, self-driving vehicles need to be as self-sufficient as possible, to avoid the hefty communications costs, network failures, and cyber vulnerability inherent in keeping the HD maps up-to-date. With the Imagry system, all data required to perform safe and accurate motion planning, based on real-time inputs, is contained within the computing platform of the vehicle. This data can be updated periodically “over the air” to benefit from the Imagry method of supervised learning.

Next stop, full autonomy!

Are you coming? Got a question for us?

    Company Locations

    Imagry, Inc.
    1630 Old Oakland Rd.
    Suite #A112
    San Jose CA 95131
    USA
    Imagry (Israel) Ltd.
    53 Derekh HaAtsma'ut
    3rd Floor
    Haifa 3303327
    Israel

    Accessibility Toolbar