The rapid evolution of smart mobility and passenger vehicles has turned advanced driver-assistance systems (ADAS) into a defining benchmark for global automotive original equipment manufacturers (OEMs). However, conventional autonomous driving architectures rely on fragmented, rule-based systems or high-definition (HD) maps, creating severe bottlenecks in scalability, geographic adaptation, and system cost. Achieving widespread mass production of urban Navigation on Autopilot (NOA) requires cost-effective, mapless AI architectures capable of navigating complex, unpredictable real-world traffic.
DeepRoute.ai, an artificial intelligence pioneer founded in Shenzhen in 2019, is revolutionizing the smart driving landscape by commercializing advanced, end-to-end Vision-Language-Action (VLA) AI software designed for commercial mass production.
Dismantling mass-production barriers
High-level urban autonomous driving has historically faced steep technical and economic hurdles that have restricted mass adoption across consumer vehicles:
- HD Map Dependency: Traditional solutions rely on expensive, frequently outdated high-definition mapping, limiting operational coverage and scalability across regions.
- Cost and Hardware Constraints: Hardware-heavy setups drive up vehicle retail prices, pricing out the mass-market consumer segment.
- Safety and Disengagement: Complex urban scenarios demand high perception accuracy and low human intervention rates to build user trust.
The DeepRoute.ai advantage
DeepRoute tackles these industry hurdles through a software-first approach that delivers top-tier performance without expensive map reliance:
- Advanced VLA Model Architecture: Integrates perception, reasoning, and control into a single foundation model, highlighted by its 40-billion-parameter VLA architecture powering DeepRoute IO 2.0. Employs chain-of-thought reasoning and real-time optical character recognition (OCR) sign interpretation for transparent, human-like decision-making and proactive defensive driving.
- Mapless and Cost-Effective Integration: Eliminates HD map reliance via map-free perception, offering flexible licensing and non-recurring engineering (NRE) models that allow OEMs to deploy urban NOA efficiently across mass-production vehicle lineups.
- Benchmark Perception and Control: Delivers market-leading perception metrics with 96% recognition and 98% detection accuracy. Averages 120 km of autonomous driving before requiring human intervention, demonstrating high system reliability and competing directly with leading EV OEMs.
Accelerating OEM commercialization
DeepRoute has rapidly established itself as a leading Tier-1 smart driving supplier, transitioning from technical validation to large-scale mass production:
- Automotive OEM Partnerships: Collaborating with major car manufacturers, including Great Wall Motor and Geely, to deploy Highway and Urban NOA across mass-production models such as Wey, Tank, and smart #5.
- Mass Fleet Deployment: Scaled its software platform across more than 400,000 vehicles on public roads1, building a self-evolving data flywheel targeting 1 million deployed vehicles for 2026.
- Universal Hardware Compatibility: Designed to adapt easily to almost any car chip, allowing automakers to integrate the software and launch new smart driving vehicles in as little as 8 months.
Back by a highly experienced technical team with deep expertise in AI vision, robotics, and vehicle engineering, DeepRoute is poised to define the future of mass-market intelligent driving. TTV is thrilled to support DeepRoute as they scale artificial general intelligence in the physical world.
