Beijing (Gasgoo)- On March 18, 2025 (Beijing time), Chinese autonomous driving solution developer DeepRoute.ai showcased its latest strategic initiative, RoadAGI, and introduced its new road-based general AI platform, AI Spark, at the NVIDIA GTC 2025.

Photo credit: DeepRoute.ai

RoadAGI marks a significant step toward achieving general AI for the physical world, aiming to equip intelligent moving agents, including autonomous vehicles, with the ability to navigate roads independently and interact deeply with their surroundings. The AI Spark platform, central to this vision, is designed to understand physical world dynamics by leveraging large volumes of high-quality, anonymized driving data. This allows AI-powered mobility solutions to move seamlessly between indoor and outdoor environments, enabling unrestricted point-to-point travel.

Since the mass production of its smart driving solutions in 2024, there had been around 40,000 intelligent vehicles sold equipped with DeepRoute.ai's technology, accumulating massive autonomous driving data. These interactions with the physical world have further refined AI Spark's capabilities in autonomous navigation and real-world interaction.

Photo credit: DeepRoute.ai

Additionally, DeepRoute.ai has integrated a Vision-Language-Navigation (VLN) model into the platform. This enables mobility agents to interpret natural language commands in real-time, correlate verbal instructions with surrounding visual data, and make autonomous navigational decisions based on their environment.

By leveraging the AI Spark platform, autonomous two-wheelers, robots, and other mobile agents can now operate beyond public roads, seamlessly navigating commercial areas and industrial parks. Furthermore, the platform enables mobility systems to function without reliance on pre-existing navigation maps, using real-time environmental awareness and dynamic obstacle interaction to determine navigation routes and facilitate unrestricted movement.

All Rights Reserved. Do not reproduce, copy and use the editorial content without permission. Contact us: autonews@gasgoo.com.