How reliable are autonomous driving systems?

19 Tem 18:51

Sensor and algorithm combos have indeed come a long way, but we're still far from 100% safety. What really stands out to me is the "false triggers" that sensors experience. For example, night vision mode triggering sensors that stop the car, or sensors in rainy conditions struggling to decide between "water droplets or an object" and activating. That's why when testing systems, you need to try them out in different lighting and weather conditions; I compare data I collect from night streets, quick maneuvers in heavy traffic, and even garage parking. As for automatic braking, it's almost perfect—I'd say it's nearly flawless. The close-proximity detection system usually responds, but due to its characteristic delay, some low-speed collision scenarios can still turn into accidents. Tesla and Honda's recent cases prove this. I think the most reliable solution is for the driver to stay alert and ready to intervene manually, rather than blindly trusting the system. No matter how advanced the sensors get, the best sensor is still a careful human eye.