The 5 Levels of Autonomous Driving ๐Ÿง  ๐Ÿš™

There is a lot of talk about level 2, 3, 5 self-driving cars, but what do these levels mean? Which level is Tesla at? What about Waymo?

Let me walk you through the different levels in this thread ๐Ÿ‘‡
Level 0๏ธโƒฃ - No Automation

Just old fashioned cars here - there are no systems assisting during driving. The human driver needs to do all the work.

โ–ช๏ธ Driving -๐Ÿง
โ–ช๏ธ Monitoring -๐Ÿง
โ–ช๏ธ Fallback -๐Ÿง
Level 1๏ธโƒฃ - Hands-on

The car takes over part of the driving.

Example: adaptive cruise control - the car accelerates and breaks automatically, but the driver needs to steer.

L1 cars have been around for more than 20 years!

โ–ช๏ธ Driving -๐Ÿง/๐Ÿš™
โ–ช๏ธ Monitoring -๐Ÿง
โ–ช๏ธ Fallback -๐Ÿง
Level 2๏ธโƒฃ - Hands-off

The car takes over both steering and acceleration/breaking, but the driver needs to be able to take over at any time.

All car companies now offer L2 systems as an option.

โ–ช๏ธ Driving - ๐Ÿš™
โ–ช๏ธ Monitoring -๐Ÿง
โ–ช๏ธ Fallback -๐Ÿง
Going from L2 to L3 ๐Ÿ“ˆ

The big difference is that the responsibility in case of an accident is transferred from the driver to the car.

This increases complexity significantly, because the car needs to handle all possible edge cases.

Regulations still don't allow L3 systems...
Level 3๏ธโƒฃ - Eyes off

The car does the driving in specific situations (e.g. on the highway). The car may hand back control to the driver, but it needs to give couple of seconds warning, because he may be distracted.

โ–ช๏ธ Driving - ๐Ÿš™
โ–ช๏ธ Monitoring - ๐Ÿš™
โ–ช๏ธ Fallback -๐Ÿง
There are not L3 cars yet.

Many companies are working on L3 systems and BMW, Mercedes and Honda already announced specific plans. However, there are still regulatory hurdles that need to be overcome.

The FSD Beta of Tesla are still a L2 system!
Level 4๏ธโƒฃ - Mind off

The car drives itself and doesn't rely on a human as a fallback - there may be no one behind the steering wheel.

However, the car can only drive in certain areas and conditions, which it supports.

โ–ช๏ธ Driving - ๐Ÿš™
โ–ช๏ธ Monitoring - ๐Ÿš™
โ–ช๏ธ Fallback -๐Ÿš™
Objective Design Domain (ODD)

These restrictions are called ODD. Example: Waymo's service in Phoenix, which is geofenced and the safety driver is present if the environmental conditions are challenging.

Another company that is allowed to drive without safety driver is Cruise.
Level 5๏ธโƒฃ - Steering Wheel Optional

The ultimate stage of self-driving - the car can handle all situations under all conditions on its own. At this stage, you don't really need a steering wheel anymore.

However, we are still quite far from this... ๐Ÿคทโ€โ™‚๏ธ
Summary ๐Ÿ

So now you know that almost all systems that are today on the market are sill L2. Yes, even Tesla...

The only comany that arguably has a true L4 system is Waymo, but it still operates in very limited situations under optimal conditions.
If you liked this thread and want to read more about self-driving cars and machine learning give me a follow! ๐Ÿ‘

I have many more threads like this planned ๐Ÿ˜ƒ

โ€ข โ€ข โ€ข

Missing some Tweet in this thread? You can try to force a refresh
ใ€€

Keep Current with Vladimir Haltakov

Vladimir Haltakov Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @haltakov

13 Apr
How Does a Self-Driving Car Work? ๐Ÿ”ง ๐Ÿง  ๐Ÿš™

This is classical self-driving car software stack. Nowadays, all steps in the pipeline are dominated by machine learning.

Read below for details on each step ๐Ÿ‘‡ Image
Sensors ๐ŸŽฅ

There are 3 main sensors for environment perception 360ยฐ around the car:
โ–ช๏ธ Cameras
โ–ช๏ธ Lidars
โ–ช๏ธ Radars

Each sensor has different advantages and disadvantages, so combining all 3 is the best strategy to achieve maximum robustness. Image
Perception ๐Ÿ–ผ๏ธ

The perception module processes all the sensor raw data to detect different objects, drivable space, lane boundaries, measure distance etc.

Fusing the information from different sensors usually increases the quality of the data significantly. Image
Read 8 tweets
12 Apr
There are different ways to detect concept drift for computer vision, depending on the application.

You will need to set up monitoring of your model in production and continuously evaluate it.

1/6
You can do spot checks on production data by labeling some small amounts and comparing the performance to the performance on the test dataset.

This is of course a very expensive and error prone approach (see Simpson's paradox).

2/6
You can evaluate some high-level statistics about the performance of your application. Usually the CV model is not the last part of the pipeline.

For example, a car recognizing speed limits may check if the driver actually drives at a speed close to the speed limit.

3/6
Read 6 tweets
22 Feb
How high can you score against my new @DilemmaBot at Prisoner's Dilemma? ๐Ÿ’ช

Just mention it in a tweet and it will start playing with you.

For now, I won't reveal the strategy it's playing, but I'll write about that later and will publish the code as well.
Here some explanation how the game of Prisoner's Dilemma works. The bot will be playing 10 rounds of it.

Read 4 tweets
17 Feb
What are Convolutional Neural Networks? ๐Ÿž๏ธ โญ๏ธ โ›ฐ๏ธ

CNNs are an important class of deep artificial neural networks that are particularly well suited for images.

If you want to learn the important concepts of CNNs and understand why they work so well, this thread is for you!

๐Ÿงต๐Ÿ‘‡
What is a CNN? ๐Ÿค”

A CNN is a deep neural network that contains at least one convolutional layer. A typical CNN has a structure like this:
โ–ช๏ธ Image as input
โ–ช๏ธ Several convolutional layers
โ–ช๏ธ Several interleaved pooling layers
โ–ช๏ธ One/more fully connected layers

Example: AlexNet
A good example - AlexNet

Throughout the thread I will be giving examples based on AlexNet - this is the net architecture that arguably started the whole deep learning revolution in computer vision!

I've written more about AlexNet here:
Read 21 tweets
15 Feb
Prisoner's Dilemma ๐Ÿค”

Time for some game theory! ๐Ÿ‘จโ€๐Ÿซ

Prisoner's Dilemma (PD) is an interesting game that explains how two rational individuals may make decisions that seem irrational.

The game has lots of examples and applications in real life!

Thread ๐Ÿ‘‡
There are different examples of PD, but this is the one I like most.

You want to buy something from another person. You exchange closed bags one containing the money and one the goods.

Both you and the other person can choose to honor the deal โœ… or to give an empty bag โŒ.
If you both honor the deal โœ… โœ… (cooperate), you both gain something.

If you both exchange empty bags โŒ โŒ (defect), at least nobody loses.

If you leave the bag empty, but get a full bag โœ… โŒ, you gain a lot, while the other person is screwed.

Image source: Wikipedia Image
Read 11 tweets
10 Feb
Dealing with imbalanced datasets ๐Ÿ โš–๏ธ ๐Ÿ˜

Real world datasets are often imbalanced - some of the classes appear much more often in your data than others.

The problem? You ML model will likely learn to only predict the dominant classes.

What can you do about it? ๐Ÿค”

Thread ๐Ÿ‘‡
Example ๐Ÿšฆ

We will be dealing with a ML model to detect traffic lights for a self-driving car ๐Ÿค–๐Ÿš—

Traffic lights are small so you will have much more parts of the image that are not traffic lights.

Furthermore, yellow lights ๐ŸŸก are much rarer than green ๐ŸŸข or red ๐Ÿ”ด.
The problem โšก

Imagine we train a model to classify the color of the traffic light. A typical distribution will be:
๐Ÿ”ด - 56%
๐ŸŸก - 3%
๐ŸŸข - 41%

So, your model can get to 97% accuracy just by learning to distinguish red from green.

How can we deal with this? ๐Ÿค”
Read 13 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!