Autobrains unveils Air2Road localization technology for automated driving
25 September 2024
Autobrains, a dveloper of AI-powered automotive solutions, unveiled Air2Road, a localization technology developed to enhance the capabilities of Advanced Driver Assistance Systems (ADAS) and Autonomous Driving solutions. Powered by Autobrains’ Liquid AI Technology, Air2Road delivers precise, real-time vehicle localization for safe and efficient navigation, even in challenging conditions such as occlusions, adverse weather, and complex urban environments.
Fully autonomous vehicles will require reliable and accurate localization to determine both vehicle position and the driving environment. Traditional localization systems, heavily reliant on Global Navigation Satellite Systems (GNSS) or crowd-sourced maps, face significant challenges such as limited coverage, high costs, and susceptibility to interference. Additionally, these systems rely on a single source of sensor-based information, leading to frequent errors and over-reliance on one type of input during driving.
Air2Road combines signatures derived from two independent modalities—aerial imagery and ground-based perception - ensuring multimodal redundancy and accuracy. These Signatures are sparse binary hyperdimensional representations of both the driving environment and objects within it. This innovative approach overcomes many of the limitations inherent in traditional systems. These include:
Multimodal Redundancy: By creating a joint representation of Signatures from two independent modalities (aerial images and ground-based data), Air2Road enables an unprecedented advantage of multimodal redundancy. Other existing systems use a fleet of vehicles with similar ground-based data sources, which increases repetitive mistakes and dependency on a single source of information.
Full coverage: Unlike GNSS-based systems, Air2Road, with its Signature-based technology, continues to function effectively even under trees, bridges, between buildings, and in any other scenario with bad reception of the GNSS signal.
Aerial Perception: Aerial images generate additional data layers—such as road markings, traffic signs, and obstacles—enabling safer and more efficient driving decisions and a redundancy to real-time perception.
Scalable and Efficient: Traditional localization systems, such as crowd-sourced maps, require a fleet of vehicles to repeatedly traverse roads for data harvesting and building the map before the localization service becomes available for any vehicle. In contrast, Air2Road, enables the first vehicle on the road to achieve precise localization instantly, ensuring scalable deployment across various regions without compromising accuracy.
Cost-Effective: Air2Road is compatible with affordable aerial imagery from multiple vendors, and its compressed Signatures technology enables large areas to be stored directly on the vehicle. This allows for a low-memory, low-bandwidth system, eliminating the need for constant cloud connectivity.
Environment-Aware Path Planning: By leveraging real-time data on road conditions, traffic flow, and parking areas, Air2Road supports safer, more optimized navigation.
Air2Road replicates mechanisms such as situational awareness and attention focus to ensure accurate decision-making. By filtering out irrelevant data and focusing on critical elements such as road edges and buildings, the system significantly improves localization precision.
Comments