DeepRoute.ai has introduced its new smart driving platform, DeepRoute IO 2.0, featuring a vision-language-action (VLA) model designed to deliver more human-like reasoning and improved safety for autonomous vehicles. The platform, which supports both LiDAR and pure vision configurations, is being positioned as a scalable solution for mass-market deployment through partnerships with five automakers.
The VLA model combines visual perception, natural language processing, and decision-making capabilities, enabling the system to analyze complex traffic scenarios with logical chain-of-thought reasoning. The company said this approach enhances transparency in the decision-making process and ensures adaptability to diverse urban driving conditions. Built on NVIDIA’s DRIVE AGX Thor platform using the Blackwell GPU architecture, the technology is optimized to handle demanding reasoning workloads required for real-time vehicle operation.
DeepRoute IO 2.0 is designed with flexibility in mind. Its multi-chip, multi-sensor architecture allows automakers to integrate the system across various vehicle types, regardless of whether they rely on LiDAR or vision-only setups. Features include optical character recognition for reading traffic signs and text-based road information, as well as a voice control interface that allows drivers to interact with the system through natural speech commands.
The company emphasized safety enhancements such as blind-spot detection, spatial awareness, and defensive driving strategies intended to reduce collision risks. A key focus is also on transparency, with the platform capable of explaining its step-by-step reasoning to reduce the “black box” perception of autonomous systems.
DeepRoute.ai confirmed that five original equipment manufacturers (OEMs) have agreed to adopt the platform, with the first vehicles expected to enter production later this year. Having been validated in real-world urban driving conditions, the technology is expected to accelerate the rollout of advanced driver assistance and fully autonomous systems in mass-produced vehicles.