Alumni Pantry Talk | Chen Nuo (MScISTM 2024) mapping the road to smarter autonomous driving
Drawing on his former experience in autonomous driving test development at XPeng Motors, Chen Nuo (MScISTM 2024) offers a frontline perspective on the real challenges behind autonomous driving — not only enabling cars to see the road, but also to understand it. Tackling three industry pain points – Simulation scenario training: Using simulation platforms such as Carla and Apollo to build scenario libraries, replicating edge cases like accidents, extreme weather or complex junctions, and identifying risks in a safe, virtual environment – Reducing trial-and-error costs: Real-world mistakes are costly. Automated testing frameworks and data replay allow defects to be detected early in the coding stage, intercepting potential collision risks – Real-time evaluation: Developing automated pipelines and real-time performance evaluation systems ensures that every algorithm update receives immediate feedback, boosting iteration efficiency The industry focus is shifting from algorithms to engineering. As algorithm gains diminish, success will belong to teams with strong engineering execution and vehicle-grade stability. Autonomous driving is rapidly permeating verticals such as low speed logistics, mining and sanitation; yet safety, edge cases, engineering execution, and regulation remain the key challenges. Soft skills: Passion for AI, with the ability to leverage generative AI and AI agents to streamline workflows; strong data sensitivity to identify systemic issues from large-scale test data; an engineering mindset with strong DevOps awareness. The office pantry is often where professionals can catch their breath amid the demands of the day. The CUHK Business School Alumni Office is delighted to present the “Alumni Pantry Talk” series, inviting alumni to share candid glimpses into their professional lives. Regardless of your industry, take a coffee break with us and enjoy our “Alumni Pantry Talk”!
The calibrator behind the autonomous driving “brain”
My work centred on processing spatio-temporal data to build and calibrate precise, real-time “mental maps” for autonomous driving systems. In essence, it ensured cars not only see the road but also understand network changes, enabling precise navigation and smooth driving.Always maintaining professional depth
I took ownership of each project, addressing testing challenges through structured tools and processes. By probing perception, prediction and planning modules, I designed scenarios that anticipate model boundaries. I continuously expanded the scenario library and aligned with frontier technologies such as VLA, shifting testing from passive fault-finding to proactive leadership.2026: From technology validation to mass production
2026 marks a turning point as autonomous driving moves from validation to mass production. With national standards for Level 3 autonomy in place, manufacturer responsibilities are now clearly defined, while “black box” systems and specialised insurance pave the way for new vehicle models.Tip: Core requirements for autonomous driving test development
Hard skills: Proficiency in Python, C++, Pytest, RobotFramework and CI/CD processes; familiarity with simulation tools (such as Carla, AirSim, or Apollo), ROS/ROS2, and Linux optimisation.