I’d like to see a competition between autonomous driving vehicles on some sort of track. Something like an AI-500. Let’s see what they can do in this kind of situation. Let’s see what that does for their autonomy.
Probably be best if a variety of manufacturers are participating in the competition.
Must have something like the flag system for signaling to the vehicles concerning course conditions. Perhaps a course boundary that shows the various signals all around the track (a yellow boundary for a yellow flag which I believe is traditionally an accident on the course).
Cars that incorporate IR (see my previous article) may have some advantages in anticipating what the other vehicles are doing on the course.
Another idea for improvements in the autonomous automobile space is the use of Generative Adversarial Networks (GAN’s) in the training thereof.
Normally in a GAN you might have one network attempting to identify birds in photographs, while the other network is doctoring photographs to get the first network to misidentify photos—either false-positives (birds where there are no birds) or false-negatives (non-identified birds).
In training autonomous vehicles, it may be possible to create virtual test environments where one network is working to successfully drive through simulations, while the second network is attempting to thwart that driving—either by causing the first network to crash its vehicle or by forcing the first network to halt its progress for trivial reasons.
I don’t know that anyone is doing this today, but it does seem a potentially fruitful avenue of testing and training.
I have been kicking this idea around for years and am only now finally publishing something about it. The title really tells it all. The next big thing? Sure. You poop in a toilet every day, so let’s make a toilet that crowd sources medical and related data to tell you how you are doing.
(I know we use toilets for urine as well, and everything I am saying here about poop should be thus extended to urine as well.)
I suppose it will need a sort of garbage disposal or blender component. It would presumably work from assays of your excrement, and make whatever various tests or analyses it will be capable of doing (which ought to increase over time).
Currently there are a number of states we can assess from human excrement. I don’t know them well enough to list them all here, but things that come quickly to mind are pregnancy, some kinds of cancer, kidney issues, bowel issues, &c.
Imagine how this smart toilet might make various important measures each time you use it. It would know you are you based on your account. It would catalog everyone in a poop database. This data could be made available to the medical community. This could be done both specifically (your doctor could be allowed to access your data) and generally (your data would be part of the big data that researchers could use to further our knowledge and advance our research projects).
It could alert and advise things such as “you should have your kidneys checked and refer to test xxxx when you do” or “you should do a pregnancy test” directly to your smart phone (or whatever method you might choose).
Now lest you think I’ve gone mad (or think this is just one more data point in a long line trending toward madness), here is a great video concerning human gut biology in which the speaker specifically mentions a smart toilet! Yay, me! (He mentions the toilet at about 50:30 but the whole talk is quite excellent.)
To my knowledge no one is currently taking advantage of camera sensors’ (typically) inherent ability to capture inferred (IR) data. If someone is, good on them. Someone will tell me at some point.
Camera sensors often (always?) are able to collect data beyond the visible spectrum into the inferred (and presumably into the ultraviolet (UV). It seems clear to me that parsing this data can be useful to an autonomous vehicle in identifying human and other animal agents on and near the roadways. (There may be similar benefits available in parsing the UV wavelengths as well.)
There are situations (especially at night) where IR will allow human and other animal agents to stand out against the background in ways the visible spectrum cannot. IR (and UV) may be able to usefully augment the lasers currently being used as well to provide a more robust dataset and thus more complete picture of the world for autonomous vehicles.
This may also help to solve some of the problems associated with windows and similar either reflective or transparent surfaces (as they should typically produce different IR ranges than the surrounding surfaces (and again perhaps UV differences)).
Another related source for potentially useful information would be comparing polarized lens data with standard lens data.
