#autonomous-driving

Tesla expanded its robotaxi service to Dallas and Houston, extending its Texas footprint while public tracker data still points to a small, closely watched fleet outside Austin.

RAD-2 reframes diffusion-based driving planners as a generator-discriminator system, then adds reinforcement learning feedback where imitation-only training is weakest. The headline number is a 56% collision-rate drop versus strong diffusion planners, plus reported real-world deployment in complex urban traffic.

NVIDIA, Hyundai Motor, and Kia said on March 16, 2026 that they are expanding their strategic partnership around autonomous driving. The collaboration links Hyundai Motor Group software-defined vehicle capabilities and fleet data with the NVIDIA DRIVE Hyperion platform for systems ranging from level 2+ assistance to level 4 robotaxis.

At CES 2026, NVIDIA said it expanded the DRIVE Hyperion ecosystem with new suppliers and sensor partners. The company positioned AGX Thor-based Hyperion and Halos safety workflows as core infrastructure for scaling level 4 autonomy.

MicroVision's sub-$200 solid-state lidar technology could dramatically shift the economics of automotive ADAS sensors, potentially making lidar viable for mainstream vehicles and reopening debate over camera-only autonomous driving strategies.