Tesla 'FSD Lite' Fails to Detect Downed Trees and Logs in Roadway

Tesla's autonomous driving ambitions faced a new round of scrutiny this week after reports emerged of the system failing to detect large obstacles in the road. The incidents highlight a persistent struggle for the company's vision-based software to handle unpredictable organic debris.

Two separate users on X posted videos demonstrating Tesla vehicles failing to recognize and avoid downed trees and logs positioned in the middle of the roadway. According to a report by InsideEVs, the vehicles continued forward despite the presence of the obstacles, creating a high risk of collision. The software involved in these instances is referred to as 'FSD Lite,' specifically FSD v14 Lite, which is a distilled version of the v14 software optimized for vehicles equipped with Hardware 3 (HW3).

The Vision-Only Gamble

These failures are rooted in Tesla's strategic decision to rely exclusively on 'Tesla Vision.' Unlike many of its competitors in the autonomous driving space, Tesla has removed radar and ultrasonic sensors from its vehicles, opting for a vision-only approach. This architecture makes the vehicle entirely dependent on a neural network's ability to process camera feeds and correctly classify objects in real-time.

When the system encounters an object that does not fit a known pattern—such as a fallen log with an irregular shape and color—the neural network may fail to categorize it as a solid obstacle. This reliance on visual classification alone means that if the software does not recognize the object as something that should be avoided, it may not trigger the necessary braking or steering maneuvers.

Critical Safety Implications

The inability to detect large, static objects represents a significant safety gap in the pursuit of full autonomy. While Tesla's systems are generally adept at recognizing standard road markers, vehicles, and pedestrians, these 'edge cases'—uncommon but dangerous scenarios—remain a primary hurdle. The failure to see a downed tree is not merely a software glitch but a fundamental challenge in how vision-only systems interpret the physical world.

For the industry, these incidents underscore the ongoing debate over sensor fusion. Critics of Tesla's approach argue that adding LiDAR or radar provides a necessary redundancy, ensuring that even if a camera fails to classify an object, the vehicle can still detect a physical mass in its path.

What to Watch

Tesla has not yet issued a formal response to these specific videos, but the incidents put pressure on the company to prove that FSD v14 Lite can maintain safety standards on older HW3 hardware. Observers will be watching for upcoming software updates to see if Tesla improves the classification of organic debris or if the system continues to struggle with non-standard road obstructions.