Development Platforms

Overview

Overview
Patents
Certifications
Partners

Air2Road

Setting localization standards for automated driving.

Utilizing Autobrains’ Signature-based AI, Air2Road matches aerial imagery with vehicle camera data to deliver highly accurate localization. The multimodal system enables centimeter-level precision, enhancing the safety of automated driving.

122 +

Of Patents Received

The Challenge

Traditional localization systems relying on GNSS or crowd-sourced maps face challenges like limited coverage, high costs, interference, and single source data inaccuracies.

The Solution

Air2Road delivers precise, real-time vehicle localization for safe navigation, even in challenging conditions such as occlusions, adverse weather, and complex urban environments.

Air2Road - Technical Highlights

Driven by Liquid AI, Air2Road focuses on features critical for localization.By matching two independent modalities, aerial imagery with 360 vehicle camera data, it achieves centimeter-level localization.

Cm-Level Accuracy

Centimeter-level precision for unparalleled accuracy and reliability.

Multimodal Sensor Input

Matching visual aerial with ground-level data inputs.

Signature Technology

Sparse binary hyperdimensional driving environment representations.

Ground View Perception

In real-time, the vehicle perceives the driving environment through a 360-degree camera ground view.

Real-time signatures while driving

The camera input is reduced to sparse binary representations (signatures), representing the driving environment and relevant objects.

Aerial image signatures

The vehicle is equipped with offline data of the aerial view of the driving environment, which is also compressed into signatures.

Real-time signature matching

Both signatures are matched in real-time, delivering accurate localization results even if one of the modality inputs is incomplete.

Precise Ego-localization (< 10 cm)

The localization is highly robust to environmental changes and offers centimeter-level (<10cm range) accuracy.

Use Cases

Air2Road Performance within Different Driving Scenarios

From bustling urban areas to rural roads, Air2Road adapts to all driving scenarios, including dense city centers and tree-covered routes.

Urban

Parking Lot

Tree-Covered

Localization in Urban setting.

Precise localization in urban environments.

Technology Used:

Self-Learned Signatures for aerial
and ground view matching.

Precise localization in parking lots.

Focusing only on localization-relevant static objects while ignoring parked and moving vehicles.

Technology Used:

Self-Learned Signatures for aerial
and ground view matching.

Stability under trees or bridges.

Ensuring reliability even in occluded scenarios and across environmental or seasonal variations in both aerial and ground inputs.

Technology Used:

Self-Learned Signatures for aerial
and ground view matching.

Full coverage

Unlike GNSS-based systems, Air2Road’s Signature-based technology operates reliably in challenging environments, such as under trees, bridges, between buildings, and in areas with poor GNSS signal.

Scalable and Efficient

Existing systems like crowd-sourced maps need a fleet for data collection before localization. Air2Road offers instant localization for the first vehicle, enabling scalable deployment without losing accuracy.

Cost- Effective

Air2Road uses affordable aerial imagery from various vendors. Its compressed Signatures tech stores large areas on the vehicle, enabling a low-memory, low-bandwidth system without continuous cloud connectivity.