Hacker News Reader

Comment by shadowgovt | Hacker News Reader

There's research on this topic, and you'd be surprised.

The average human apparently has a 2.3 second delay to unexpected, interruptive stimulus while driving (https://copradar.com/redlight/factors/IEA2000_ABS51.pdf). We almost never perceive it as such because we tend to measure our own reaction times from the point we are conscious of stimulus to the point we take willful action to respond to it, but the hard numbers appear to show that critical information can take 1+ seconds to percolate to conscious observation (remember, the brain spends a lot of time figuring out what in the soup of sensory nonsense is worthy of attention).

View on HN

The critical part is that you need to compare apples to apples - in this case, the one second delay is from the point at which the car had a clear idea of there being an obstacle in its path until it would have started to apply the brakes. If you want to compare this to humans, you also need to remove the sequence of time during which the human identifies the potential obstacle as relevant and subsequently as something he would crash into.

Whether this time is shorter or longer for humans is another question entirely (though the human intelligences' ability to deduce intent from behavior and forecast actions of other humans in traffic should give robocars a good challenge in that department as well). But in terms of raw reaction time after determining "I have to brake NOW", a human is definitely faster than one second.

That study shows a ~1 second delay from the incursion being visible to a human to them releasing the accelerator and a further 2.3 seconds to have the brake fully depressed, by which time they were also steering. The study also implies the average human response time was adequate to avoid the collision...

To put things into perspective, the Uber spend 4 seconds after actually detecting the incursion trying to figure out whether it needed to respond and doing nothing, then a further second of enforced pause after concluding it did need to do something until finally starting to reduce speed 0.2 seconds before impact.

That study doesn't purport to be an objective measurement of absolute reaction time. It's comparing relative driver behavior between a driving simulator and a test track, and it doesn't seem to have controlled how immediately drivers needed to respond to avoid a collision. It does, however, include one objective measurement of human reaction time, albeit not as a primary result of the study:

> the time from incursion start to throttle release included the reaction time of the tow vehicle driver pulling the foam car (which was consistently less than 200 milliseconds)

Reply on news.ycombinator.com