DiDi Autonomous Driving starts R2 robotaxi trials in Beijing and Guangzhou

Why it matters

Bob Bo Zhang is pairing DiDi's ride-hailing distribution with GAC Aion's manufacturing to move R2 into passenger service. Missing fleet and intervention data still leave scale and economics unclear.

DiDi Autonomous Driving began fully driverless passenger trials with its R2 robotaxi in selected areas of Beijing and Guangzhou on August 31st, moving the vehicle from road testing into bookable rides through the DiDi app.

The launch puts Bob Bo Zhang, DiDi Autonomous Driving's co-founder and CEO, behind a vehicle designed to connect three pieces that robotaxi developers often have to assemble separately: autonomous-driving software, automotive manufacturing and an existing ride-dispatch network. Zhang has led the autonomous-driving operation since it became an independent DiDi unit in August 2019. Before DiDi, he held a senior technology role at Baidu, and his graduate work at the Chinese Academy of Sciences focused on human-computer interaction and artificial intelligence.

R2 was jointly developed with GAC Aion, the electric-vehicle arm of Guangzhou Automobile Group. According to DiDi's August 31st announcement, the vehicle carries 33 sensors, DiDi's full-stack Level 4 driving system and a central computing platform that combines three vehicle domains. DiDi also built in multilayer hardware and software redundancy.

The release establishes a limited trial, not an unrestricted commercial service. DiDi did not publish the number of R2 vehicles in operation, the size of the service areas, fares, passenger volume, autonomous miles or the frequency of remote assistance and human intervention. Those figures will determine whether R2 is becoming a transportation service or remains a tightly controlled demonstration.

Zhang's distribution bet

DiDi's advantage starts before an R2 reaches the curb. Riders can request one through the same app that already handles human-driven trips, while DiDi can apply its dispatch, mapping and fleet-management systems to autonomous vehicles.

That operating model has long shaped Zhang's program. DiDi Global's 2025 annual report said autonomous driving would become commercially viable through integration with an existing mobility platform. The filing also said the operation used traffic data from DiDi's shared-mobility network for localization, prediction and vehicle control. DiDi has not disclosed its current autonomous-driving headcount.

The filing was equally direct about the remaining distance: DiDi Global said its autonomous-driving technology had not been commercialized at scale as of the end of 2025.

The August 31st step puts R2 into passenger trials and expands its use to selected parts of Beijing. DiDi says R2 was delivered in January 2026 and has since been deployed for public-road testing in Guangzhou, Beijing, Shenzhen and other cities.

A robotaxi designed around the back seat

R2's cabin shows how DiDi is applying its ride-hailing experience to the vehicle itself. The robotaxi has enlarged rear seating, high-angle adjustable seats and a 17.3-inch ceiling-mounted display. Riders can use voice commands to verify a trip PIN, start the ride, adjust the air conditioning, open windows and change the ambient lighting.

DiDi says an R2 can open its windows before pickup to ventilate the cabin. Planned additions include an exterior communication screen, temperature and lighting presets, and tools that help passengers find the vehicle through horn and voice prompts.

These details address mundane problems that become operationally important when no driver is present. A robotaxi has to identify the correct passenger, explain when it is ready to move, resolve pickup confusion and give riders control of the cabin without help from the front seat. Zhang's human-computer interaction background is particularly relevant here: R2 has to make a driverless trip legible to passengers, not simply navigate the road.

DiDi says R2 meets five-star standards under C-NCAP and Euro NCAP, although the announcement does not cite independent test reports or detailed ratings. DiDi also claims more than 2,200 consecutive days of safe autonomous operations through its Huiju Port operations center. That number is a calendar streak and provides no mileage, ride or intervention denominator.

GAC is financing the manufacturing bridge

GAC's role goes beyond supplying a vehicle platform. In October 2024, GAC Group led DiDi Autonomous Driving's $298 million Series C, with DiDi participating. The financing was earmarked for autonomous-driving research and production of the jointly developed robotaxi.

DiDi Autonomous Driving followed with a 2 billion yuan Series D in October 2025, worth about $275 million at the time. Participants included DiDi, GAC Group, Zhongguancun Science City Technology Growth Fund, Beijing Information Industry Development Fund, Beijing Artificial Intelligence Industry Investment Fund and Guangzhou Guanghua Fund. DiDi Autonomous Driving has not published a valuation for either round.

That investor lineup reflects the scale of Zhang's assignment. Level 4 software has to survive urban traffic, while the vehicles must be manufactured, maintained, cleaned, charged and positioned economically enough to compete with human-driven rides. GAC gives DiDi a manufacturing partner and a supply chain. DiDi contributes demand, dispatch and daily operating infrastructure.

Competitors are already publishing the scale metrics that DiDi omitted from the R2 announcement. Pony.ai reported that its robotaxi fleet had surpassed 1,400 vehicles, alongside fare revenue and per-vehicle order data. DiDi's app distribution could become a substantial advantage, but R2's fleet size and utilization will show whether that distribution is producing repeatable autonomous rides.

For Zhang, the immediate achievement is getting a purpose-built vehicle into passenger service. The next proof points sit outside the cabin: fleet growth, intervention rates, paid demand and the cost of keeping each R2 on the road.