L'IA peut-elle conduire une voiture de manière autonome sur autoroute et dans la circulation suburbaine à grande échelle ?

Waymo a ouvert son service commercial entièrement sans conducteur au public à Phoenix en octobre 2020 — la première fois que le public pouvait rouler sans conducteur de sécurité.

Background

Waymo launched its fully driverless commercial taxi service in Phoenix, Arizona, in October 2020, becoming the first company to offer public rides without a human safety driver onboard. AI-driven autonomous systems have demonstrated capability in structured environments such as highways with clear lane markings and minimal disruption. However, suburban traffic introduces added complexity due to unpredictable elements like unmarked roads, varied signage, cyclists, pedestrians, school zones, and construction detours. Multiple companies—including Waymo, Cruise, Zoox, and others—have conducted trials and demonstrated autonomous navigation in suburban areas, but none have achieved routine, large-scale deployment. Current systems typically fuse inputs from cameras, lidar, radar, ultrasonic sensors, high-definition maps, and deep learning models to interpret surroundings and plan trajectories. Despite advances, edge cases—such as erratic behavior from other road users or rare environmental conditions—continue to produce failures in safety-critical scenarios. Regulatory uncertainty, public skepticism, and the need for robust fail-safe mechanisms remain significant barriers. Ongoing research focuses on improving sensor fusion, prediction accuracy, and scalability across diverse geographies and weather conditions. As of May 9, 2026, autonomous driving at scale is still an active area of development and testing, with no systemic solution yet proven under all real-world conditions.