IA developed by MIT that would further improve autonomous driving

Source:  hibridosyelectricos.com 

Greetings comrades, this week I bring a post that relates the IA with the automotive world. As you know there is already IA applied in the so-called autonomous vehicles.

The one I come to comment on in my post was one that has developed the people of MIT, since this was presented in 2017 at the International Conference on Robots and Intelligent Systems (IROS), but only now is that they are being tested and has proven to be effective even more above the one used by many autonomous cars today which is called LiDAR, this consists of sensors that examined a field of vision and made decisions based on what was learned.

Very good and everything, but there are always better options in the world of technology, as everything is updated and improved respectively. In this case the weak point of the LiDAR system is only take into account the image available in the already mentioned "field of view".

So everything that was outside of it would be invisible so to speak, so that under certain conditions the LiDAR might not react to such a scenario. 

Source: st2.depositphotos.com  

The people at MIT decided to solve this problem, since it is common to always see in closed streets, as well as in parking lots, and among a great variety of places, such as people "appear" or a car from nowhere, and that is who has not gone through the situation for example in a parking lot in which you are in the task of getting a place to leave the car and in that walk, there is a child running out of the cars, or because you are not in a street and out of nowhere a ball quickly crosses in front of you, as they say "behind a ball there will always be a child" and 2+2 are 4 so there are situations to consider to improve.

And that's just to give some examples, as I said before everything advances and the AI applied to autonomous cars is no less... So the solution to this at MIT has called it ShadowCamera.

The developers of this are the CSAIL (Computer Science and Artificial Intelligence Laboratory), if the same mentioned in one of my previous posts.

ShadowCamera works as its name indicates with the shadows, specifically those that are located on the ground, for this the system uses a set of cameras that then transmits these to the IA and this in turn is responsible for detecting small variations found in the images and then makes use of a "artificial vision" to calculate if there is something approaching us so this around the corner and thus avoid the possible collision. 

As I told you, the project has been going on for a few years now and it is that at the beginning (in its first version) some tags were implemented to align the area with the area of artificial vision, but this was not so practical in the day to day so they gradually evolved to what it is today, making use of visual odometry (used in the Mars Rover if that NASA robot).

When they tested the ShadowCam, installed in an autonomous car, they made use of a parking lot under certain conditions such as their lights turned off (in order to imitate night conditions). He noticed that the car revolved around the pillars 0.72s faster than the LiDAR system. Also thanks to adjustments made for such conditions the system achieved an accuracy of about 86% accuracy.

It is hoped that the development of this system will continue and that in the future the ShadowCam system can be used regardless of the environment in which it is located. And it also applies to conveyor robots and motorized wheelchairs.

Personally I think that this type of technologies are very profitable and to measures that these can be perfected or joined with other already existing ones, as for example it occurs to me because not to join this one with the one that mentions in my post of that IA capable of detecting us in spite of not being visible, it would be a very but very complete system.

Little by little everything revolves around AI, so we will be more and more involved in it on a lesser or greater scale, but as death sooner or later will come.

With nothing more to say thank you for getting here and reading my post, I'm not leaving without wishing you a happy day, late or night.

Links of interest 

MIT News  

Fuente: Mi galería de Giphy