Researchers conducted an extensive real-world study using a Tesla Model Y equipped with Full Self-Driving (FSD) software over more than 100 days on Queensland roads in Australia.

Across this period, they recorded over 500 safety-critical events requiring human intervention or revealing significant limitations in the system’s perception and decision-making abilities.

While the technology demonstrated impressive capabilities in many routine driving scenarios—often matching or exceeding human precision—it also showed consistent weaknesses in common and complex traffic environments.

Key problem areas included confusion at small bridges, misinterpretation of time-restricted speed zones such as school areas, and unsafe handling near railway crossings where the system failed to adequately anticipate risk.Complex urban conditions such as roundabouts, poorly marked streets, and interactions with cyclists or e-scooter users also frequently led to errors.Weather conditions that obscured lane markings further reduced system reliability.

Notably, none of the test journeys were completed without at least one human intervention, highlighting that current systems are not yet fully autonomous in real-world Australian conditions.The study argues that the challenge is not solely technological but also infrastructural.

Many road features designed for human interpretation—such as inconsistent signage, faded markings, and ambiguous intersections—contribute to machine errors.

The researchers suggest that improving road infrastructure, standardising signage, and enhancing lane clarity could significantly improve autonomous vehicle performance.

They also propose a complementary role for autonomous vehicles as mobile sensing platforms capable of detecting potholes, damaged signage, and congestion patterns in real time.

Rather than viewing autonomous vehicles as entirely independent systems, the study advocates for a hybrid approach in which smarter vehicles and smarter infrastructure evolve together.

The findings highlight that widespread deployment of autonomous driving in Australia will require coordinated improvements in both vehicle intelligence and road system design.