[On the Road] "Ten Times Safer Than Humans"... Pony AI Robotaxi Fleet to Reach 3,000 by Year-End

Visiting Pony AI's Beijing Research Center

Experiencing the Robotaxi Firsthand, Seamless Ride

Accumulated Mileage Surpasses 80 Million Kilometers

"Our Robotaxi is fully driverless and has been developed so that, even if one sensor group fails, the other groups continue to operate reliably. We estimate it is about ten times safer than a human driver."

RoboTaxi exhibited at Pony.ai Beijing Research Center. Photo by Youngchan Choi

On April 23, I visited the Pony.ai Research Center located in the Yizhuang Economic and Technological Development Zone in Beijing, China. Throughout the explanation of the history and technological achievements of Pony.ai, the pride on the staff's face was evident. Pony.ai is a Chinese autonomous driving company that was listed on the NASDAQ in the United States two years ago, and on the Hong Kong Stock Exchange last year. Currently operating 1,100 Robotaxis, Pony.ai plans to run 3,000 Robotaxis in the global market by the end of this year.

After touring the research center, I had the opportunity to ride in a Pony.ai Robotaxi. Sitting in the passenger seat, I fastened my seatbelt and pressed the 'Start' button, and the car began its journey to the destination. Through the map interface, I could see nearby vehicles, pedestrians, and structures in real time. The Robotaxi even recognized traffic lights hundreds of meters ahead and displayed how many seconds remained until the signal changed.

During the test ride covering just over 10 kilometers, not a single dangerous situation occurred. The driving was smoother and safer than that of a human driver, allowing me to browse for music and other entertainment on the central screen instead of watching the road.

The true value of the Pony.ai Robotaxi became apparent on narrow alleys and busy roads. Parked cars lined the right side of the road, making the street extremely narrow, while jaywalking pedestrians, motorcycles, and bicycles constantly appeared on the road. Despite these challenges, there was not a single sudden stop or abrupt acceleration. The vehicle gently decelerated and skillfully avoided obstacles.

Central screen of Pony AI Robotaxi. Photo by Youngchan Choi

Although the Robotaxi could be described as nearly flawless, it was also fully equipped with safety options for emergencies. Each Pony.ai Robotaxi is monitored from a central control center, where a human operator oversees between 10 and 30 vehicles. When the Robotaxi determines it needs human assistance, it requests help, at which point the operator suggests a route. The Robotaxi then decides independently whether to follow the suggested route. A representative from Pony.ai explained, "Most of the time, we simply monitor the screens. If the Robotaxi doesn't request assistance, there is no need to intervene."

Operational efficiency in maintenance has also been achieved. After completing its journey, the Robotaxi returns to the garage and parks itself automatically. Then, a single staff member is able to clean, charge, and service 20 vehicles within an hour.

The service quality as experienced by customers was also highly satisfactory. When the reporter got out at the wrong destination, the Robotaxi automatically rolled down its window, enabling communication with a control center staff member via the central screen. If a passenger left belongings behind, an audio alert instructed them to check for forgotten items before exiting, and the vehicle would remain stationary until the items were retrieved.

Pony.ai has already accumulated 80 million kilometers in total mileage through its global transportation operations. This means that its autonomous driving technology continues to improve as more data is collected every day. As the fleet size grows from the current 1,100 vehicles to 3,000, the rate of mileage accumulation will inevitably increase. With the 7th generation of vehicles, Pony.ai has reduced the size of the LiDAR sensor, resulting in a more natural exterior appearance. Although the number of sensors has increased compared to the previous generation, the manufacturing cost of the hardware has been reduced by approximately 70%.

This content was produced with the assistance of AI translation services.