In an era when autonomous driving is highly dependent on digital and electrical infrastructure, Waymo’s recent experience during a large-scale power outage in San Francisco unexpectedly became a real-world test of how driverless systems respond to extreme scenarios. The outage, triggered by a fire at a power substation, disrupted the daily lives of up to 130,000 customers and directly impacted Waymo’s autonomous ride-hailing service. Multiple vehicles came to a halt mid-journey at intersections or in the middle of roadways, causing tangible disruptions to urban traffic and the passenger experience.

Level 4 autonomous driving has already been achieved by some automakers and activated on specific routes, but mass-market personal adoption is still some distance away. Tensor’s Robocar, promoted as the world’s first personally owned Level 4 autonomous vehicle, raises the question of whether it can redefine the future of the automotive industry in terms of technology, business models, and market positioning.

Chinese electric vehicle manufacturer Xpeng recently unveiled a series of new technologies at its Tech Day, including a self-driving taxi, plans for a flying car business, and a next-generation robot. However, the centerpiece of the event was undoubtedly the upgraded semi-autonomous driving platform “VLA 2.0,” which has drawn significant attention across the automotive industry. The system will not only be rolled out across mainland China but will also be made available for other automakers to adopt. German automotive giant Volkswagen will be the first partner to collaborate on this technology. Xpeng made it clear that the goal of VLA 2.0 is to surpass Tesla’s Full Self-Driving (FSD) system.

Most mainstream autonomous driving and ADAS (Advanced Driver Assistance Systems) architectures today rely on a “modular sensor fusion” approach, in which multiple cameras, radars, and LiDAR sensors collect data that are processed through several algorithmic layers for object recognition and environmental modeling, followed by decision-making and control. While this structure offers high controllability, it also entails lengthy development processes, high labeling costs, and accumulated errors between modules that often lead to less natural and consistent driving responses. Honda Motor Co., Ltd. recently announced an additional investment in California-based AI startup Helm.ai, aiming to shift its autonomous driving and ADAS development strategy from “sensor-oriented” to a new era of “AI End-to-End (E2E)” learning architecture.

Chinese startup DeepRoute.ai recently unveiled its next-generation driver-assistance platform, DeepRoute IO 2.0, featuring its self-developed VLA (Vision-Language-Action) model. The company claims this architecture breaks through the interpretability and generalization bottlenecks of traditional end-to-end (E2E) systems. The new platform integrates visual perception, semantic understanding, and action decision-making, offering functions such as spatial semantic comprehension, irregular obstacle detection, traffic sign text recognition, and memory-based voice control. This delivers a more humanlike and safer advanced driver-assistance experience. Compared with current market solutions, DeepRoute IO 2.0 offers greater technical transparency and stronger mass-production adaptability, having secured five design-win projects and approaching 100,000 vehicles in deployment. The company also plans to accelerate passenger car applications and expand into Robotaxi operations as part of its roadmap toward Road AGI (Artificial General Intelligence for Roads).

Among the hottest topics in today’s industry, the application of AI across various fields is revealing the potential to revolutionize existing practices. In the field of intelligent transportation, AI’s potential is particularly vast. At the ongoing ITS World Congress in Atlanta, USA, Sony Semiconductor Solutions is showcasing two smart city pilot projects that demonstrate how its advanced edge AI sensing technology helps local governments improve traffic safety, enhance roadway efficiency, and support infrastructure decisions with high-precision, long-term data—creating new opportunities for both the automotive and transportation sectors.

As autonomous driving technology edges closer to commercial deployment, U.S. ride-hailing platform Lyft has announced the launch of its inaugural Driver Autonomous Forum, a new initiative specifically designed for seasoned drivers on the platform. The goal is to give human drivers a voice in the transition process as robotaxis are gradually integrated into Lyft’s service system, allowing them to participate in decision-making and influence future strategy.