DeepRoute launches IO 4.0 with complex spatial driving

When vehicles meet on a narrow road, one of them must first make room before either can continue forward. On August 31, DeepRoute.ai released DeepRoute IO 4.0, an OTA (over-the-air software update) version for production vehicles. It adds system-level understanding of the spatial relationships between vehicles and their surroundings, allowing the system to autonomously plan reversing to escape construction zones and actively reverse to yield when vehicles meet on narrow roads.

The capability also covers pulling away from roadside parking spaces and multi-point turns. When space is limited both in front and behind, the vehicle can maneuver back and forth; when its original direction of travel is blocked, the system looks for a new way through based on the surrounding space. DeepRoute describes this as expanding assisted driving from a focus on driving scenarios to complex spatial control covering both forward and reverse directions, and says it is the first company in the industry to systemically integrate this capability into an assisted-driving system.

IO 4.0 completed optimization of 12 assisted-driving capabilities. Data released by DeepRoute shows that perception and recognition improved by 41.7%, dynamic interaction improved by 38.6% and driving efficiency improved by 53.6%. The system also adds comfortable and standard driving styles, and adjusts its handling of complex intersections, merging and lane changes, dynamic traffic-rule understanding and interaction with vulnerable road users; starting at traffic lights and vehicle control over speed bumps were also included in the optimization of the everyday driving experience.

****However, the relevant features will be rolled out successively according to notifications from partner automakers, and unified details on specific vehicle configurations and timing were not provided at this launch. The improvement percentages above come from DeepRoute’s own data.

Behind this production upgrade is DeepRoute’s further investment in physical AI—enabling artificial intelligence to understand and act on the real world. In May this year, the company established Superfluid Lab, researching foundation models for the physical world and exploring VLA, world models and multimodal approaches. DeepRoute views IO 4.0 as a product path for solving specific problems on real roads, while using foundation-model research to explore the limits of spatial understanding and capabilities in complex, open environments.