Xpeng has outlined a central part of its latest intelligent driving tech and autonomous drive that is powered by AI.

It says its upcoming VLA 2.0 (Vision-Language-Action) architecture is the next-generation AI foundation for intelligent driving.

By streamlining how visual input translates into vehicle actions, Xpeng claims the new system is designed to deliver faster response times, reduced information loss and more human-like driving performance across complex real-world scenarios.

Building on this technology, Xpeng says it is advancing its robotaxi roadmap in China, with trial operations planned for later this year.

Xpeng has also showcased its AI-driven ADAS system to global delegates in China. It has hosted live road demonstrations during the UN/WP.29 Informal Working Group on Automated Driving Systems (IWG ADS) in Shanghai. The international forum, which brings together stakeholders from regulators to industry experts and consumer groups to develop global harmonised rules for Automated Driving Systems (ADS), convened offline in China for the first time last week.

The IWG ADS primarily invites government and industry representatives to contribute technical expertise. Xpeng has participated in the Taskforce on Advanced Driver Assistance Systems (TF ADAS) discussions on Driver Control Assistance Systems (DCAS) since 2023 and began joining IWG ADS meetings in 2025. At the Shanghai session, it says it was the only emerging Chinese automaker to participate throughout and provide live demonstrations to global delegates.

Officials and experts from major automotive markets including Canada, the European Union, Japan, United Kingdom and the United States, experienced Xpeng’s system in real-world urban and highway scenarios, marking ‘one of the most direct engagements yet between global regulators and a Chinese smart EV manufacturer deploying AI-driven ADAS at scale’.

Xpeng claims the session provided global delegates with ‘firsthand exposure to how large-scale, data-driven AI systems perform in the variability and unpredictability of real traffic environments’.

During the demonstrations, participants observed the XNGP driving system’s real-time perception, decision-making and control capabilities, as well as its integrated safety framework — including driver status monitoring, human-machine interaction logic and safety design. Across complex urban and highway scenarios, these capabilities translated into a smooth driving experience, the company says.