Driver Drowsiness/Inattention Detection



Detect when the driver is becoming drowsy to alert the driver, or possibly take over if Full Self Driving is available.


There are various ways to detect this, from cameras pointed at the driver, to how poorly a person is driving. Here’s more on detection methods.


Driver monitoring is anticipated for later in 2021.  New inside cameras added to the mid-year 2021 S and X, and internal software is already written to monitor the driver.

edited by moderator
Category: CY3XS Applies to:
     Created 25-Nov-2017


I agree that it is one of the driver's key responsibilities.  It seems to me that a more useful question to ask is, Can technology assist the driver in upholding that responsibility?  Drowsiness detection, could be an excellent driver assist.  There is evidence that a significant cause of driver accidents are the following, among them drowsiness:

Many of these appear to be detectable using the data provided by the car and its significant AI resources.  Having the car monitor a driver for apparently unsafe conditions if the driver asks for it seems to me to be a reasonable assist.   I even agree that such detection may be subjective.  However if you consider that subjectivity means that there is a distribution of probabilities that the subject is indeed "drowsy" if they show the characteristics of "drowsy"; then it means that you can select the acceptable probability before invoking a driver alert.  Allowing the driver to have control over this setting seems to me to be a fair way of adjudicating subjectivity.  I suspect that this approach could save driver lives and prevent accidents.
    Created 30-Mar-2018
Deciding that someone is "too" sleepy sounds pretty subjective.  This is one of the driver's key responsibilities.
    Created 30-Mar-2018
Since there is a camera inside the TM3, it should be possible to train the AI to recognize the characteristics of a sleepy driver along with his/her driving characteristics to obtain a likelihood of drowsiness (or even inebriation).  Using this data and a settable likelihood threshold (possibly user settable) to invoke a stimulus-response protocol.  For example,  "Hey, <profile-name>, you seem to be drowsy.  Are you awake?  If so please tell me how many dogs you see on the screen."  No response causes the car to slow down and pull over.  An incorrect response can  lead to further actions including contacting 911.  Presumably there should be a limited number of car initiated responses and a clear responsibility line demarcation.  I would propose something like the priority would be protection of immediate danger to the driver/passengers and then to the car.
    Created 18-Mar-2018
One of the best ideas I've seen here.

I foolishly drove home from Palm Springs recently when I was much too tired. I only woke up when I drifted out of lane... or when I was honked at. When you realise you're too tired, it's too late. Teslas have all the necessary hardware for most of the likely detection methods (wiring electrodes to the driver is a non-starter).
    Created 1-Dec-2017