#China #Guangzhou #Autonomous Driving #Robotaxi #Pony.ai

Pony.ai’s Breakthrough in Autonomous Driving

Pony.ai, a frontrunner in the autonomous driving sector, recently celebrated a significant milestone: its Level 4 (L4) automotive-grade autonomous driving domain controller equipped in the latest seventh-generation Robotaxi fleet has achieved over 2 million kilometers of on-road testing. This achievement highlights the system's remarkable hardware reliability and software adaptability, setting a new benchmark in the realm of autonomous vehicles.

Unveiled at the Shanghai Auto Show in April, Pony.ai's L4 domain controller was crafted entirely with automotive-grade components, designed to operate for a decade or 600,000 kilometers. The entire autonomous driving stack, which includes the domain controller, signifies a leap forward in terms of performance, durability, and cost-effectiveness. Notably, Pony.ai managed to slash the domain controller's cost by 80% compared to its predecessor, thanks to extensive algorithm optimizations and advanced system engineering.

The cutting-edge domain controller, developed entirely in-house, stands as a world-first L4 automotive-grade controller. It employs three NVIDIA OrinX chips in its main unit and a fourth chip to ensure system redundancy, culminating in a staggering 1,016 TOPS of computing power. Pony.ai's innovative approach merges hardware with software development to meet the stringent functional and operational standards of fully driverless Robotaxis and also to satisfy the performance and safety metrics required for L4 autonomous driving algorithms. This results in significant reductions in size, weight, power consumption, and costs—optimizing hardware expenditures considerably.

This central computing hub integrates diverse sensors and components necessary for Robotaxi operations, such as power management systems, interactive displays, hazard lights, gateways, and Global Navigation Satellite System (GNSS). By consolidating these features into a single unit, Pony.ai has successfully simplified wiring complexity and minimized the number of components required, leading to lower manufacturing costs and enhanced vehicle layout efficiency. This integration accelerates assembly line processes, paving the way for more rapid mass production. The controller also supports both liquid and passive cooling, ensuring continued operation even in the face of potential cooling system failures.

A key feature of the system is its multi-layered safety architecture, which employs degradation strategies for maximizing operational safety. In instances of primary system failure, the controller can switch to either the redundant system or the Minimum Risk Condition Controller (MRCC), ensuring continued safe vehicle control. This capability allows the vehicle to navigate complex situations such as intersections while continuing to maintain crucial perceptual functions.

At the heart of Pony.ai's system is a steadfast commitment to safety, achievable through full-vehicle redundancy and highly intelligent algorithms. Utilizing the PonyWorld foundation model and reinforcement learning, the company has consistently delivered driving performance that exceeds human capabilities and continues to advance its system's ability to adapt to real-world complexities.

As Pony.ai’s seventh-generation Robotaxis enter mass production, they are simultaneously undergoing rigorous road testing. This combination of robust engineering and innovative design promises to vastly expand Pony.ai’s fleet of driverless vehicles while enhancing user mobility experiences. With 2025 marking its first year of mass production, the company aims to deploy a fleet of 1,000 autonomous vehicles, a vital move towards achieving operational efficiency and economies of scale.