AI & ML in the Automotive Industry #machinelearning #ai

Key Takeaways

Hello and welcome everyone to another session in the data art series. We are thrilled to be here with you this evening for a session full of action learning. I'm Ashosh part of the data science team at analytics with for those who have joined us for the first time a brief introduction to the data art session. The data arc is a series of webinars conducted by analytics vidia and led by top industry experts. It is a fun way to understand the concepts of data science from the leading players in the data tech domain and as the name suggests it is it's one arc dedicated to data. We are hopeful that these sessions are going to be a great source of enrichment and value adding for community members. Now onto our session today which is AI and ML in the automotive industry. The use of AI ML in automotive industries has increased rapidly. Now industries are using AI and ML in all the major processes. In this data, Abhishek Raj Permani will talk about many such use cases like self-driving car, predictive maintenance, improved safety and many more. He will also explain the different advanced driver assistant system features which are being used in current cars at the different levels of autonomous cars. I hope you're excited to attend this data with us. Before we kick things off and I hand it over to our speaker, a quick recap of the housekeeping items. First, we are recording the session and we'll make the recording available in a few days on our YouTube channel. Second, please use the Q&A section for asking any questions you might have during the session and we will do our best to answer them as the desire progresses or towards the end. Third also we will share a poll about the feedback of the session towards the end of the session which I request you all to kindly fill up. Now on to a speaker in this session of data we have Abhishek Raj Pahani with us. Abishek is currently working as a machine learning engineer with Jaguar Land Rover. He's a computer science graduate from IIT. In the past he has worked with Samsung research and development institute India in the data science team. Till now he has presented three data science papers in international conferences. He is not only a researcher but also a mentor to more than 2,000 students through his data structure and algorithm courses and various live sessions. He has always given back to the community. Over to you Abishek. The virtual stage is all yours. Thank you Ash. Thank you very much for this introduction and for having me on the session. Okay. So okay I'll start my sharing. Can um can you allow me to share my screen? I am unable to do that. Okay. Until the uh all the thing will do. Uh I think now I Okay. Cool. Hello guys. Okay. So before starting to the session what we can do is uh it would be great if you can uh mention your name or which year you are from or what background you come from or which company you are currently working in if you are already passed out. So just mention all these thing in the chat so that I can get an idea what type of uh crowd I am uh we have today. Okay. So yeah we'll also answer all your questions. So don't need to worry about that. I can see. Yeah. Okay. Great. Let me shift it to here. Okay. Now I can Yeah. Hi Capel. Hi. Hi Chandra. Hello. Okay. Cool. Cool guys. And yeah from electrical okay most most of you seems to be MSE or uh a statistic students MT student as well in data science okay MSE data science and computing okay great okay okay this seems good and uh let me change position of my screens. It will start that. Okay. So, it might look that I am uh talking I am looking into another screen and my camera is here. Okay. But my PPD is open on this side and I'm my camera is on this side. Okay. So, kindly bear with that and Okay. Entrepreneur from Iran. Okay. Great. Hello everyone. Okay. So, let's start the session then we have everything set. Okay. So, let me know in the chat uh uh if you if my screen is visible and then we can start. Yeah. Okay. Visible. Yeah. Cool. Okay then. So, I hope my screen with presentation is visible, right? anyone? Yeah. Okay. Good. So, here we are with the very interesting and very uh topic which is very close to me uh from last 1 1.5 years and uh so let's see what we'll cover in this session. So it will be uh I will give you an idea an overview about uh what is artificial intelligence what is machine learning and uh what is autonomous this automotive industry. So we'll go through the basics from that and we'll not go much in detail in artificial intellig machine learning because they analytics with already has sessions on that and that are very good. So I have also seen few of them so they are very good. So you can go through them uh for going in detail of that and uh then we'll come to more detail on the automotive side. Uh that's what because that's because I am here and uh what lies ahead of us like in the automotive industry what is ahead for us and there is some amazing fact that most of you and I I was also not aware about when I joined. Okay. So what is artificial intelligence? So uh it's basically the ability of a computer to think and learn. That's the book definition and that I found on Wikipedia on some platform I guess. So basically it's an ability of computer to think and learn. As simple as that. Uh there are three parts involved. One is learning. Uh another one is reasoning and another one is uh selfcorrection. So in terms of learning we have taking taking and acquiring data and then creating some algorithms out of that data. Okay. The algorithms can be of uh many types you have uh you have seen you you know many of the algorithms of machine learning. So like regression or uh some computer vision algorithms. So all those algorithms we can use on that data and we can get the uh supposed output proposed output. Okay. So create learning in is that phase then reasoning comes then choosing the right algorithm. Now there are many algorithm like linear regression multiple all those regressions algorithms are there clustering algorithms are there. So which al which algorithm is suited for which type of data that you need to figure out or uh which algorithm will give the best output that you require. Okay. So that we need to check out that comes in the reasoning part. And third one is the selfcorrection that is finetuning of the parameters of the algorithm. to give the more accurate result. So that is the part of selfcorrection. Okay. So uh yeah so that is a overview idea of what artificial intelligence is. So let me tell you how the session will go and how I like to take the sessions generally. Uh so uh I start with the session uh and we'll take we'll take the content what I have prepared and once that is done I have in the end the Q&A part as well. So when that is done, uh you can Okay. Okay. Cool. I'll switch to that. Okay. Now I think it's good, right? Why this is coming? No, I can't switch to slide mode because it would cover this is maximum I can do because it would cover uh it is covering my other slide as well. That is problematic for me because I can't see your questions on that. So uh during the session as well you can write your questions in the chat and I will take that if that is related to that current slide. I will take that start taking that question or if that that doesn't disturb our flow. Okay. But uh I will take the question in the end. If you have some question you can uh write in the chat and we'll take it the end. So no need to worry about that. Okay. So that was an idea of artificial intelligence. Now comes machine learning. So many people are confused about machine learning and artificial intelligence. I was also confused in my college days. But now uh when you deep dive into what machine learning artificial intelligence is uh through various sources you will find that artificial intelligence is the subset of the machine sorry machine learning is the subset of artificial intelligence that uses statistical methods. Okay. Uh again it can be of uh regression. It can be of uh some clustering methods or algorithm that it uses to improve the results basically on the training data. Okay. So now the types of machine learning are so AI is a bigger picture and machine learning is a smaller picture compared to that and in machine learning we have a subset called deep learning okay so that's how it works AI is a bigger picture then machine learning and then deep learning so that's how it works and machine learning now the types of machine learning are supervised unsupervised and reinforcement supervised uh where we use the labelled data okay where the data is correctly labeled like uh let's say there is a data of cats and dogs. So that's very famous example you all of you have must have heard that. So let's a data of cats and dogs and it is clearly mentioned that this fe this picture belongs to cat and this picture belongs to dog. So that algorithm can clearly identify okay this is cat and this is dog and it can retain or it can create some features uh the algorithm can create some features. Okay cat has these types of nose, cat has these types of like triangle-shaped uh ears or something. it it develops some features so that it can identify or distinguish between cats and dogs. So that's the part of machine learning and that algorithm creation comes under what we saw earlier that is reasoning part okay or learning part you can say like reasoning is then again picking up the algorithm okay so that's how it works and so supervises having the labeled data okay and unsupervised is having the unlabelled data and the unsupervised learning algorithms will make kind try to make some clusters out of it okay it will try to make some clusters out of it so that uh you can clearly clearly identify what are cats and what are dogs. But again the data given to supervised learning algorithms are not labeled. Okay. It is it would it won't be clearly mentioned that these are cats and these are dogs. Okay. Okay. Reinforcement learning. Now reinforcement learning has something to do with uh uh rewards and penalty system. So let's say uh you are a uh let let so in the reinforcement learning uh let me consider myself as an agent and everything other than me is a involvement. Okay. So let's say I'm taking this session. So I'm doing an action by taking a taking this session. So I may get rewarded or penalized based upon my action based upon my taken taken action. Okay. So that's how the reinforcement learning works. There is a system, there is an agent and everything other than agent is a it's its environment and agent does some action on the environment and it takes the agent to some other state. Okay, while moving towards other state it has taken an action. Now this action can have penalty or reward. Getting my point? It has penalty or reward. So if it has reward now the system will learn or the algorithm will learn okay this is a good action and I can do this. So, so that's because I will get rewarded or if it is a wrong action, it will get penalized. Let me give you an example of self-driving car. It's not how completely the self-driving car works, but to give you an idea of what reinforcement learning is. So let's say the self-driving car is moving on the road and there is a pedestrian in front of it on the road and let's say self-driving car in the initial stage hits the pedestrian and now this is the action taken by the self-driving car. So consider self-driving car as an agent. Now it has it hit the pedestrian and now pedest now this action taken by the self-driving car will be penalized will be the algorithm of the self-driving car will be penalized for this action. Obviously it will penalize. So now in the future when the self-driving car sees a pedestrian it will not hit the pedestrian it will remember okay I in yes uh before I was penalized for this taking this action so it will either stop the car or it would uh turn change it lane or go other side okay again it depends upon the algorithm but uh getting a point reward and penalize system is what makes the algorithm better and better so that's where the reinforcement learning comes okay So I hope you got the idea about AI and ML and what's the difference between AI and ML. So everything that is covered in machine learning is a part of artificial intelligence that is in short and machine learning has something to do with statistical methods which is uh training of the data and then uh putting some algorithm on the data and getting the proper output. So there are three types of uh or two types of data that we majorly deal with uh that is one is train data and second one is test data. So train data is when we so when we get the data okay so when we get the data out of it let's say we get a data of let's say 10 lakh 10 lakh let's say vehicle images we get okay so we'll keep 40% of this data for the training set and 60% for the test set again this uh ratio or this uh proportion depends upon algorithm to algorithm and organization to organ or organization or again use case to use case for that matter. So 40% and 60% is taken separately and on the 40% of the data you will train the algorithm. Training means you will test the algorithm you will develop the algorithm you you the algorithm during this stage will develop the features will develop the parameters and all those things. the algorithm will get ready and you will choose the algorithm on the training by training on the data and once that is done you will test the algorithm on this test data because this test data is something which algorithm has not seen so far. So that will be very helpful for algorithm to fine-tune and to make minute minute changes so that it can get better to the unknown data as well because ultimately what your aim is to give an better algorithm for the unknown data not for the known data. Okay, every time in real life you won't face situation that is already seen by you. So that's where the point lies for an algorithm. Okay, so training data and test data that's the two things. So these are the basics of AI and ML. So so that uh if anyone of you don't uh have idea about AI and ML, this is what AI and ML looks like. Okay. Now quickly jump into the automotive industry. What automotive industry is? You must have cars or you must have seen cars at least. The automotive industry includes industry associated with production, wholesaling, retailing, maintenance of motor vehicles. Motor vehicles can be uh two wheelers and four-wheelers. Okay. So again we in the session will mostly focus on four wheelers that is cars. Okay. So uh can you guess in the chat uh what's the approximate line of code that you think uh will a modern car have? I'm not talking about the completely self-driving car that is still a dream for many companies. But a normal modern car has which has a good infotainment system and uh good features based on the software. What what you think will have lines of code uh you can predict in like like numbers like 10,000 lines of code or millions, billions, 100 billion, 10 million something like that 10 million. Okay, a wild guess would be enough and believe me if you guess if you're guessing then you will be uh getting a good answer out of that. Okay, so many zeros to learn. Okay, others what do you think? 10 million 100k. Okay, 10k thousand side of code. Okay. Any other wild guesses? Okay. So, uh I'll paste a link in the chat. Okay. And here you go. So, I think this is visible to everyone. Yeah, everyone it is going to Yeah. So this is the um if it is connected to real life analysis it can be a few lines. Okay. Okay. Let's see how many lines. So uh basically if you go up uh Google Chrome has this 10 million. This is in millions. Okay. So Google Chrome has 10 million. the source is clearly visible uh which I'm dealing with. Uh so Google Chrome has around 10 million line of code and Linux kernel has around uh let's say five or 6 million lines of code. Okay. And uh what you say Apache open office has around 20 20 around 2022 million lines of code. Boeing uh plane has total flight software. This is Boeing 787 has around 15 billion line of code. So this is very amazing when I uh when I was researching for the this particular topic and uh like how many lines of code do cars have this this was very shocking for me as well because we have not uh exposed to that amount of data because you can't see that amount I mean you are getting my point you can't see that data on your daily system and daily work right I can't I see I can't see that many lines of code on my daily basis so the only source is internet And Microsoft Office has around so Microsoft Office complete tool of 2001 has around 25 million line of code. Okay. And Windows 7 completely has around 40 millions lines of code. Window XP that you uses Windows 7 that we uses previously has this. Microsoft or Mac has this lines of code. Facebook if you see Facebook has around 60 million line of code. 60 61 millions lines of code. And when you come to the car software, it has around 100 million lines of code. So that was very shocking. And this car, this car software is average modern high-end car. It's not completely autonomous car. Okay, that is still a dream as I mentioned and that is under development. So and this data this uh complete stats is of from 2017 and from 2017 the car industry has the automotive industry has gone like anything. So in 2017 according to that data and that level of uh autonomous car or modern car you can say has around 100 millions line of code you can imagine the lines of code the car goes into and then uh so that's the I have pasted the link in the chat you can go through in your after the session or whenever you like okay so that's the num that's the amount of lines of code the cars this. Okay. So, coming back to our topic. Yeah. So, uh I was totally amazed by the lines of code the cars has and then we consider that only the IT industry has the lines of code but you can see the automotive industry has so many lines of code to work on. Okay. So, yeah. Coming on to next uh where automotive industry uses AI and ML. This is direct onpoint topic where automotive industry uses AI and ML. So first thing that everyone comes into everyone's mind is uh autonomous vehicles right partial or fully autonomous. So let me tell you a thing like uh there are five level of auto autonomous vehicles level one to level five and currently what Tesla is building is level three or level four something around that level of autonomous vehicles and level five signifies that the vehicle would be completely autonomous. It would it won't require any human intervention from traveling from point A to point B. So number of so the number of companies that are working on in the automotive industry and on the autonomous vehicle and electric vehicles are like Apple is working Google Whimo has Google has its project Whimo working on and they are still running on I think New York street and they are using Jaguar card only for testing their uh some feature they are I was reading about the other day. So Google is using Google is developing their cars. Apple is in the car market now. Microsoft softares most of the companies are using Microsoft Azure or other softwares. Nvidia is already there. So mobile companies like Oppo, Xiaomi, all these are also building their cars and getting into the automotive industry. So the automotive industry is now like booming any day because uh very much people have said very uh known people have said like last 10 years was for the IT industry and IT industry has seen its peak now and now the next decade belongs to the automotive industry because auto automotive industry has already be already been a bit behind the IT because the product cycle of auto automotive industries is in years compared to the IT companies which has a shorter product cycle. Okay. So anything which has which is current which is currently implemented will come in the cars in the future. Okay. So that's where the gap lies between IT and automotive industry. So that gap is filled by today's software companies and partnering with automotive country companies. So if you are working for an IT company, if you want to work for an IT company, it it there are high very high chances that you might uh end up working on driverless cars or EVs or everything. Okay. Because if you're working for Nvidia, Nvidia has this this team separate team working on that. Okay. So with the testing underway by companies including Tesla and Ford and the UK arrivals and Berlin based web fully autonomous vehicles will likely become a reality in years to come. Okay. While testing an infrastructure and legisl legislation that is the main problem the industry will face in a completely uh driverless future or future is uh likely decades away as it says it is quoted on the internet. Okay. And lower level of self-driving technology are already widespread as you can see. So basically the cars we drive at this point like you may take L i10 or Hyundai or the normal cars we drive on the road the consumer vehicles uh we drive they are also having some L1 feature. So L1 is very basic where do you don't have any autonomous kind of feature you just have some basic infotainment system or some lines of code that uh that that would help in the all all the software like you can say it's it's like u you you now the cars are getting buttons. Okay, the remote you are getting for the car. So all those features have some lines of code. So there is L1 level and L2 level what we drive. So Tesla is building L3 and L4 level. So L5 level is something which we want every automo automotive company wants to achieve in the self-driving space. Okay. So features some features like uh lane assistance and self parking and Tesla's autopilot are example of existing autonomous uh vehicle technology. So that was for the fully autonomous vehicles. And now if you move towards the electric and low emission vehicles. Uh AI is also helping uh engineers to develop the next generation of uh electric and low emission vehicles. And thanks to machine learning models uh that can rapidly predict how batteries will respond under different condition and engineers are uh iterating on fast charging technology. So there has been a lot of research uh going by other companies like Apple is investing, Nvidia is investing so much heavily in the universities and in the companies uh on the batteries itself. The research is going on on the how to use AI and ML in batteries and in all those vehicles is a lot of research and all all a lot of money has been invested in and so fast charging is another space we are looking at and so intelligent charging infrastructure will help drivers of the future ensure that they will never run out of the power. So that is uh the if you have seen the example of uh I think cyber trucks was there by Tesla uh where if you apply brakes it will convert the kinetic energy to the electrical energy and something like that. So it will other it will somewhat charge the battery itself when and driver applies a brake on the cyber truck or the I think the trucks developed by the Tesla. I was going through some video other day. So that's the research amount of research is going on to help the automotive companies and vehicle design and testing it's another feature it's another domain that AI and ML is useful to the automotive industry like traditional vehicle design and testing is costly and time consuming endeavor so especially when significant problems aren't identified okay uh until after a physical vehicle is built okay so until we find a very physical damage Uh so vehicle design and testing are very difficult task in that but uh computer modeling has typically been used to play out specific scenarios with artificial intelligence for cars. Automotive engineers can model uh that perfectly mirror every aspect that mirrors the every aspect of the vehicle design and test the vehicle under realistic and dynamic scenarios long before it is already been built. Okay. So these all uh computer modeling systems and uh using AI and ML I don't know much in the detail of the modeling aspect because the kind of work I do is more on the self-driving cars and for the what you say on the camera module and all those things. Okay. So but there is a lot of research going in that direction and a INML is helping a lot that now again manufacturing equipments is something while register assembly line like robots are not examples of AI new AI application being used on the manufacturing floor are revolutionaries revolutionarize how cars are produced okay like BMW for example uses AI powered robots to build uh to build custom cars and autonomously transport their materials transport their materials while avoiding uh moving objects and people okay so that's where they are cutting their cost as well as using the AI and ML to transport vehicles and autonomously do all these things okay now since these tools are powered by neural networks again a type of AI technology they continue to learn from their environment That's how their algorithm is built on and allowing them to adapt more quickly to challenges that arise. So that how it is helping in manufacturing equipment. And now moving on to the quality control part. Since no manufacturing system is perfect as you know it's essentially to it's essential to quickly uh identify components and the cars that don't uh meet brand standards. Okay. And uh Audi uses computer vision to identify uh correct parts. So that is very interesting. You you can see Audi uses its crack it uses computer vision systems to identify the cracks. Okay. So that's where the things are getting interested nowadays and poss uses uh AI during testing to identify noises. So there is a separate team called aostics that work on uh that has the data of audios. Okay. And AI can also identify detects up to 90% more efficiently than human because already you know there are human errors and there are limitation to human behavior and testing and their sense. Okay. But for AI and ML they keep on learning from humans and from uh their involvement and the challenges they face. Okay. So that's where the automotive industry is benefited by the use of AI and ML. So I think yeah now let us see what ADAS means. So most of you how many of you have uh heard about this term ADAS? Quickly uh let's do a survey in the chat. How many of you have done this? Yeah. No. like uh if if I want to name some few cars like XUV7O has some ads features and uh there are I don't remember other cars in the consumer vehicle like all those Jaguars and Land Rovers and all those car have ads features but I don't remember the names of the consumer vehicle low-end consumer vehicle that has ads features yeah Tesla and all those higher stand high high-end cars have these features but XUV 7O is I have I have seen some videos of XV7's having adaptive cruise control and I think navigation system is also there and autonomic uh like MG cars has this automatic emergency braking system as well. So MG and XCB7O if you get the time after the session you can Google it and you will find a lot of videos. So so let's come back to the slide and what ADAS means. So almost all vehicle accidents caused by the human error which can be avoided by uh so basically ADS stands for advanced driver assistance systems and the role of ADS is to prevent the deaths and injuries by reducing the number of car accidents and the serious impact of those that can cannot be avoided by the like human make errors and ads systems can help them to reduce these kind of errors. Okay. So let us understand each term one by one. So all these are the examples adaptive cruise control, automatic parking and navigation system and automatic emergency braking. So all those are uh example of ads feature. There are many tons of ads feature that uh at this point automotive industry uh provides to the consumers. Okay. So there are many autonomous features like ADS features sorry and so let us understand one by one. Some of them are very obvious and by its name like automated parking it uh helps to inform divers of unseen areas. Uh so that so they know when to turn the steering and when to stop. Okay. And Audi has that feature. I have seen that Audi has that feature. You can Google it and you will find that. And vehicles equip so vehicles are basically equipped with uh rear view cameras and uh have a better view of their surrounding than the traditional side mirrors. Okay. And some systems can even complete parking automatically uh without driver's help by combining the input of multiple sensors. So basically ADAS uh involve the use of sensors and cameras. So there are many types of sensors, there are many types of cameras. Okay. So it involve the use of sensors and cameras. So one thing you all might have seen uh I have also the wagon I have at my home uh there is a system when we are taking like in we are in the reverse case and we are taking back we are taking car at the back and there is something on the back side and the car starts it's the beep sound in the car starts or the distance is calculated in Boleno as well uh in wagonar it has feature and in new i10 I think it also has this feature that the beep sound starts And there is a distance calculation that is been so it is all it is an ID feature and uh I think you have seen that right guys anyone experienced that? Anyone has experienced that feature while taking back in the car? No. Yes. Yes. Okay. Cool. So that feature comes under ads as well. Okay. It's not mentioned here but it comes under a because anything that uses sensor and cameras and which is uh which reduces uh which reduces the chances of that and injuries and accidents in the car that comes under ads basically. So yeah and automatic parking it's obvious by the name it helps the drivers using the cameras to do that and uh navigation system car navigation systems uh provide onscreen instructions voice and voice prompts to help the driver follow a route while concentrating on the road. So on the dashboard on above on the dashboard there is a wind on the what to say on the mirror you will see the navigation system now nowadays it's printed I don't have this image but you can see the above the dashboard you have this navigation system uh will be displayed on the screen so now nowadays it's changing and changing because uh while you are driving the car you don't you have to see on the outside and you can't focus on the road so now the companies provided it on the above the dashboard it is print it it will get displayed the navigation system will get displayed on the above the dashboard so that you can focus on the road and you can see the navigation system as well so that's where the innovation and research is going on so automatic emergency braking again it uh automatic it uses sensors to detect whether a driver is in process of hitting another vehicle or the other objects on the road so it automatically applies the brake Okay. So, MG has that feature. I know that and I think XV70 also has that feature. Need to confirm that though. And uh adaptive cruise control is very important feature. So, let's say you are driving on the highway. So, basically what happens in the drive on the highway if you have if you have driven there like the cars are very far like comparatively far there is comparatively less traffic on the highways. So you need to uh drive for so long period and uh so it's very difficult to monitor their speed and other cars over a long period of time. Getting my points and the accidents are very sudden. Okay. And let's say you there the road is completely empty and you want to just relax, sit back and relax while driving the car continuously. So what you can do is uh this adaptive cruise control uh control uh can automatically accelerate and slow down and at times to stop the vehicle depending on the actions and other objects in the intermediate area. So you will set uh send set a particular distance that you need from the car in the front let's say 50 m 100 m you can set and the and you can leave the steering as it is and the adaptive cruise control will uh change the lane some in some cases and can also lower uh and and increase the speed depending upon the distance from the front vehicle you have set. Okay. So you can you can put your hands down on from the steering and you can relax. But again uh these uh having these features can sometimes be very dangerous because again these are not 100% we are not in the L5 level of auto autonomous vehicles. So that can be sometime dangerous but again it's a feature so we can use that. So I hope ADAS was clear for you like what ADAS is and uh what all things are covered in ADAS and you get an idea now you might have some topic to discuss with your friends and with your colleagues that okay you know this is the adash feature and this is not the adash feature and you know what is adash feature. So now it's becoming very common and it's very good to see that people's people now understanding what ADAS is. And uh I was I was talking to one of my friends who was not aware uh we generally don't talk on the automotive side. I don't we don't talk on the work but uh he come comes up with okay Bish you know what ADAS is. ADAS has this this this. So I was I was working I I was I am working in the ADAS team itself in Jer and uh so I were very surprised now the common people also know what ADAS is. So it's a very industrycentric term but now the common people thanks to all those companies like Rata Motors Mahindra and all those companies who are bringing those things up in the market. So common people are also knowing this term. So it's very good to see that India is progressing on this side. So yeah, so let's move on to what lies ahead has a ahead of us. So uh this is the number of car that will be fully autonomous by 2030 and as you can see the growth is exponential and uh all the big car companies are all the big IT companies as well are investing in autonomous cars like Tesla is obviously there and Google is there, Apple is there, so Apple cars are there. So there are many surveys going on. So there was one survey that people would like to take a Apple car than buying a Tesla car. So there was one survey on the internet I found the other day. So that's the thing. So Apple is entering into that market. I think Apple is uh collaborating with Kia I think on this on this I'm not sure though is collaborating with Kia for that. And so yeah this is this is the new vehicle market share of autonomous vehicle. So by 2030 it would be something around as you can see it would be fully autonomous or conditionally autonomous again conditionally autonomous now you can uh now you know that what are the level of autonomicity like it's L1 L2 L3 L4 and L5 so mostly autonomous vehicle are from L3 L4 it starts from L3 L4 that like Tesla is L3 L4 level autonomous okay and L5 the Tesla is building on and working on so L3 and L4 the main problem So many companies including all those companies like luxury brands as well like Mercedes, Audi and all those things uh all those companies has L3 L4 level of autonomous cars but the main problem with them or with the industry complete is that we are having a limited amount of data and the rules and regulations that are very important to run on the roads the autonomous vehicle are very limited and they have not been said properly. Okay. Like uh the other day I was reading an article the Tesla car collides with some uh showroom get into the showroom the autopilot mode. It was on autopilot mode and it killed some people. So that was the news in China. It was also the case. I have seen that video. So it's very horrifying like how these things are working. But uh again so that's the problem lies like it can be only the problem is of testing. You have the algorithm, you have the data, train data, but now what is the test data? The test data is the real life. Okay, you have the test data but again it is limited facing the real life situations like u uh there is an there was a news I was reading like Tesla uh the Tesla car identifies sun or sun I think it was sun or yellow moon I guess uh on the evening it identifies that and it basically uh detect it as a yellow light and it slows the vehicle slows down so that is now very dangerous like vehicle is slowing down on a normal road and it identifies it as a traffic signal. Okay, yellow traffic signal. So, it slows down the vehicle. So, now this is very uh what to say now this is the lack of uh the test data or the accuracy you can say. So, this accuracy term is very important uh in our automotive industry. So, that's where many companies and all those companies are struggling with. So we need a real life data but we can't test it on the road because it is very dangerous for humans right because it is not fully L5 level but again L5 level needs to be tested on the road but we can't test it on the road because we have rules and regulations for that and it can be very dangerous consequences for human so that's the problem uh all those uh the industry is figuring out then again COVID came and the supply chain was disturbed for all those companies and due to semiconductor shortage and supply chain were disturbed now supply chain in supply chain We also use AI and ML and that is helping a lot to regain the supply chain and to getting it more smooth. So this is what lies ahead of us. Now amazing fact was there was a company there is a company sorry there is a company called Carvana. Uh it is an online uh online used car retailer store in Arizona and the company is fastest growing online car used dealer in the United States and it is known for its multi-story car vending machine. So you must have seen vending machines for like chips or cans like Pepsi can, Coca-Cola cans, all those things. We have seen vending machines in our offices and in public places. But there is a vending machine for the cars. That is something which I was not uh aware of few months back. So but when I read this article so it was very amazing that there is a there is a company who is building who is building the vending machine for the cars and you can literally if you go on the YouTube and you type uh carvana a vending machine and you will see the video there is a coin basically there is a coin I can't show it on the here because I think it might get copyright uh so there is a you will get a coin and then when you put the coin in the vending machine there is a door in the vending machine you when you put the coin uh it detects and it detects it gets inside just like we order on something from the vending machine the car will come slowly come down and on the exit it will the car will come down automatically so that's how the vending machine of the car works so it's it was very interesting and hope hopefully someday we'll experience that in real life so yeah I think that was all from my side and uh yeah thank you and I'm open for questions if you have and Yeah. Yeah. I to take uh carrier ahead on automotive industry a IML. Okay. Cool. Sorry to disturb you. But before we proceed to answer the questions, I would like to request attendees to please fill in the poll about feedback as it help us to conduct more sessions. Okay. Yeah, guys, please fill the poll. Uh they are the analytics video is working like anything. I'm I'm really glad that uh to connect with the analytics community. Okay. So yeah, how to take career uh ahead of ahead on automotive IML. Okay. Capil. So if you search on LinkedIn or if you search on Google like how many companies uh how how many companies are there okay how many companies are there who are working in automotive industry and on the software side as well so you will get Nvidia you will get Google is also hiring for software engineers who want to work on the automotive car or self-driving cars so that's the point so how to take forward is learn a IML learn data science There is a lot of scope of data science and a IMF because I have been taking sessions on AI and ML and data science nowadays. So you can learn all those things like sources are available. Panel with is one such a great source. They have very good blogs and everything. So once you get the idea you can directly apply to the uh automotive companies. Okay, that's the thing like again there are few things which they might require that is very industry specific but again uh they consider all those things. Okay, so but a IML knowledge of a IML would be must I think and again uh I have worked in MATLAB as well. So if you know MATLAB if you are someone who is working in MATLAB you have the chance you there are uh like there is a system engineering domain as well where you have where we use uh MATLAB. Okay. So you have that chance as well. So you can if you are someone working on MATLAB, you can apply as then as well. So just go to LinkedIn and search for uh companies autonomous companies which are auto automotive companies that hire for software engineers. Uh let if I want to name few that would be Jaguar Land Rover is there. Uh Mercedes-Benz is there and I think Rolls-Royce also hire from India. I'm talking about India specifically. Uh Audi also hires. BMW group also hires and Aptive is there. So there are luxury brands and there are some companies which provide services to all those companies. Okay. So Nvidia that are very important. Nvidia is there, Nvidia is there and then active is there and uh Harmon is there and uh there are other companies I'm not remembering there much names but again you can if you go through uh if I have that link handy I'll share in the end there are hundreds of companies which are hiring uh automotive companies which are hiring in India for software engineers and all those domains. So you can look into that as well you can directly Google and you will find the list. Yeah. Is mechanical engineering and data science a good combo considering automotive field is a future trend? Yeah, definitely it's it would be a great combo actually because you have the idea of the cars and mechanical side as well and you have the uh domain knowledge of data science. So it would be really great because now you have the practical idea of the field and you have the software that you can use that practical idea. So it would be great one day. Yeah. Thank you Sep. Thank you. Thank you John. Thank you Aditi. Yeah we can connect on LinkedIn or you can directly mail if you have anything um any questions if if it was if you are not getting at this point you can directly mail me or you can connect me on LinkedIn. Directly search my name on Google and you will find the LinkedIn ID. identify the edge detection and the neighboring pixels and sun traffic lights it'll get properly identified. Yeah. Yeah. Yeah. That's what they are working on because the Tesla hires very good engineers. They hire from PhDs. They hire from very good engineers. Okay. So that's what they are working on. They have they know all these things they have. But again the accuracy and real life exposure because uh sometimes on the road you will face some situations which is not the part of the training data or the test data itself. So on the real on the road you face such situations. Okay. So again the algorithm need to learn that. So yeah. Okay. I think uh sleeping during features to take snap of sleeping or Yeah. Yeah. Yeah. There is uh I think uh drows drowsiness feature uh lazy drowsiness or something feature like that on the autonomous car itself or I think it's a feature of ads. We were I was also working on this. So it detects the also I was working on like it detects the red redness of the eyes or some something so that it can uh directly ask the driver to consult a doctor or to go to uh or to take rest something like that because again it is the part which can create accidents. So you just anything which can create uh accident with and the we can create a feature out of that. Yeah. where you go with the LinkedIn profile of mine. Okay. Is predator maintenance is time series. Okay. Uh not exactly time series. Audi to okay ashto. It's not exactly time series but again somewhat related to that. I've not worked much on predictive maintenance so can't comment much on that in detail but you can Google it and you will definitely find it but now nowadays tons of data is available on internet for automotive industry earlier u few years back it was not the case now everyone most many of them are working for auto automotive industry there are many many benefits of working for automotive industry Okay, great. Then guys, uh do poll uh do complete your poll and how you like the session, how you like the PPT uh and we can they can organize more session and they are they have been organizing sessions a good amount number of sessions. Yeah, do support them. Anything else anyone wants to ask be grateful you can go on. So thanks a lot Abhishek on behalf of analytics today I would like to thank you for your time and for delivering such a wonderful session. I'm sure our audience found it insightful and hopefully we can conduct more such sessions with you in the future. Yeah sure would be love would love to do that. Yeah I hope you guys have filled a feedback poll. If not, I request you to please fill the poll and if you wish to conduct a webinar or are facing difficulty in registering, connect with us. The recording of the session will be available in a date on our YouTube channel. Okay. So, we'll be back with another session of the data on September 20. The link is in the chat section. Till then, bye-bye and keep learning. Thank you. Thank you.