This is a summary of the full article in The Conversation:
We took a self-driving car on the road for 100 days to see how it handled itself – these are the mistakes it kept making
The article describes a 100-day real-world test of a Tesla Model Y using its Full Self-Driving (FSD) system on Queensland roads. During the trial, researchers recorded more than 500 safety-critical events where the driver had to intervene because the system made mistakes or was unable to handle the situation safely. Despite the technology performing impressively in many routine driving scenarios, it never completed a single trip without requiring human intervention.
The researchers found that the system consistently struggled with everyday driving situations rather than rare “edge cases.” Common problems included:
- Misinterpreting road markings and unusual road layouts (such as narrow bridges).
- Incorrectly applying time-based speed limits, particularly school zones.
- Difficulty negotiating railway crossings, zipper merges, and complex roundabouts.
- Confusion on poorly marked or steep streets with parked cars.
- Misclassifying e-scooter riders as pedestrians.
- Reduced performance in poor weather when lane markings were obscured.
The authors argue that the findings show autonomous driving technology is both more capable and less capable than many people believe. While it can drive smoothly and accurately much of the time, it still struggles with situations that experienced human drivers handle intuitively.
Rather than expecting vehicles alone to solve these challenges, the researchers recommend making roads more “machine-friendly” by improving lane markings, signage, intersection design, and speed-limit information. They also suggest autonomous vehicles could help road authorities by acting as mobile sensors that detect potholes, damaged signs, and other infrastructure issues.
The key takeaway is that preparing for autonomous vehicles requires improvements to both vehicle technology and road infrastructure. The authors believe governments and industry should prioritise fixing the common, repeatable situations that currently cause problems, rather than focusing only on rare edge cases.
Zuduo Zheng. (2026, July 3). We took a self-driving car on the road for 100 days to see how it handled itself – these are the mistakes it kept making. The Conversation. https://theconversation.com/we-took-a-self-driving-car-on-the-road-for-100-days-to-see-how-it-handled-itself-these-are-the-mistakes-it-kept-making-286265

