Features

In conversation with Ritukar Vijay from The Hi-Tech Robotic Systemz (www.hitechroboticsystemz.com)

Novus Drive
We speak with Ritukar Vijay, Head of Technology, HiTech Robotic Systemz Pvt. Ltd. about autonomous driving

Interview by Afzal Rawuther

Q. How long have you spent in this industry. It has been quite a long and exciting time, right?

I have been in the robotics and autonomous vehicles for the past 10+ years It was like a sci-fi and the only tech that was known to India was what the IT companies were doing but today we are beyond that in the age of AI & Machine learning making real life changing technologies. The Hi-Tech Robotic Systemz started back in 2004-05 when our founder Anuj Kapuria dropped out of his PhD at Carnegie Mellon University to pursue his entrepreneurial journey.

Anuj Kapuria, during his time at CMU was part of the 2004 DARPA challenge. His classmate Chris Urmson, who was instrumental in creating Google Self Driving Initiative (now called Waymo) was also a part of the challenge. All this research and development is being pioneered by the same set of people and we have learned from the experience.

Our focus was on developing the core technology and getting the appropriate intellectual property rights. Nowadays its difficult to get those rights, owing to the number of players involved as today it is more evolved or rather evolving technology areas. Initially the only clients for us were defence research labs. We did a few path-breaking projects for them, such as state of the art unmanned vehicles and then we explored multiple verticals – Autonomous Mobile Robots for warehousing & manufacturing facilities and finally mainstream automotive OEMs. Today we are working with 7-8 OEMs (both domestic and international) providing them with factory fitted ADAS and fully Autonomous Vehicle dev roadmaps.

Q. Hitech has been at the forefront of the autonomous driving. It started working in the industry when the industry didn’t exist in India. What is the Indian automotive industry’s preparedness for autonomous driving?

When you talk about the Indian automotive industry, or rather global automotive OEMs they are not prepared with the pace of innovation happening in the space. Be it shared mobility, ADAS, autonomous vehicles, which is a big reason for recent multi billion-dollar acquisitions of silicon valley startups. It is because everyone is desperate to get their hands on autonomous technology and in order to not get disrupted.
In the Indian scenario too, the same thing is happening, where most of the OEMs are not prepared for the change and also the western tech doesn’t work here because of the great change in driving conditions in India. What is not supporting the Indian automotive industry is however, the infrastructure. It is not helping in autonomous driving coming to India anytime soon.

Q. The Hi-Tech Robotics recently came up with a fleet of self driving buses called Novus drive for large campuses. Was that an easier problem to solve since it was in a semi-controlled environment or was there any other reason to start with a problem like Novus drive?

Novus drive is basically a core software platform, which is agnostic to any vehicle, be it cars or buses. This particular use case for autonomous buses was our way of demonstrating the technology without any legal hassles. There are no legal issues with testing autonomous vehicles inside campuses. If we were to test it on roads, we would need prior approval from the government, which we are in the process of getting. We need to make the government understand that India has to be at the forefront when it comes to autonomous driving. We are trying to communicate that aspect to the government. So, the Novus drive on campuses was easier not in terms of technology but in terms of the legality. We did ferry people from the gates to the VIP lounge and various lounges at the Delhi Auto Expo 2016 using vehicles with Novus drive enabled shuttle and the vehicle performed well despite of the traffic and a lot of pedestrians. When it comes to India, it is a very chaotic environment and making the vehicle perform in such conditions, is a problem we believed we should solve.

Q. What is the biggest obstacle to widespread adoption of autonomous driving right now? Is it financial restrictions or something else?

I think money is not a deterrent right now, and that’s applicable for all the technologies. The smartphone that you and me are carrying right now, would have costed 5x if not more. The technology would be cheaper if it is widely adopted by the market. It is more about the legal issues surrounding the safety about having completely autonomous vehicles on the road. We have a different take on it ,  that it is a gradual transformation. For example, more than 75% cars in India doesn’t have automatic transmission, they still have manual gear transmission and that will be an incremental change to reach a point of some autonomy. We are filling these gaps by the Advance Driver Assistance Systems (ADAS). So, we have created products that can go into the cars today giving warnings, such as forward collision warning, pedestrian warning, and identifying different types of vehicles / road users. These systems can be scaled up to Autopilot mode along with Automotive OEM’s roadmap. Under the hood these products are processing a lot of data using machine learning with which more features can be added later on. For instance, we are working with one of the manufacturers in India for the forward collision warning systems, it will be soon a part of their vehicle itself but that’s not all, data will be collected for various other scenarios like emergency braking and traffic signal detection and other critical features for driver safety. The emergency braking feature enables  the vehicle to brake itself anticipating a collision; a traffic assist feature, where in the vehicle follows the vehicle in front of it during slow moving traffic. These are the things that can be customized for various Indian traffic scenarios. Over thye years our tech overcame challenges such as – variety of driving scenarios, traffic situation, driving style and overall mindset.

Q. What type of sensors and cameras is Hitech currently working on? Is it radar, lidar, ultrasonic? What combination is Hitech currently working on?

For all our ADAS systems, we work on optical sensors, which are camera systems with multiple cameras, where they are creating the stereo images just like our eyes and that is superimposed on the monocular classification view to perceive depth, our monocular camera feed is classifying the objects like cyclists, motorcycles and pedestrians and also estimating the distance using machine learning with a proprietary neural network. This helps us to make predictions about the objects in front of the vehicle and that helps us to take critical actions and also customize it for different scenarios, for instance the implementation would be different in a car and in a bus.

Q. For the system that you were talking about there has to be some supervised learning involved. Some test videos and some test images would be used, I believe.

It uses deep learning along with traditional computer vision techniques. It is a bit of a fusion of these technologies. We have a proprietary deep learning neural network. This is special because we are not just dependent on the amount of data. Apart from data learning we are using the traditionally computer vision techniques to identify the objects in the images. It’s a hybrid architecture, but going forward we will certainly need a huge amount of data to train it specifically for covering the variability in the Indian driving  scenarios, such as cities, small towns, villages, among others

Q. Is there an emphasis on using sensors that can be accommodated at a lower cost or are all efforts now focused at bringing the best technology available, no matter the cost?

Cost is obviously on our minds right now.  And that’s  the reason why we are not using expensive sensors for our ADAS offerings, such as 3D LiDARs. For a lot of our early projects, we were using high priced lidar for fully autonomous driving. For large adoption of ADAS, we are maintaining sensitivity about the cost to the OEMs. That is why 85 percent of the sensory perception would be camera vision. Camera sensors we believe would be most prominent.  However, if there is a radar, our system can work along with it too and enhance the long distance capability.

Q. When do you expect to see Autonomous cars on Indian roads?

I will be very blunt. The next 4-5 years will be important and safety will be of utmost importance. We are creating a model wherein Level 0-3 cars are ready for all India roads. Level 3 will be exclusively available only for the highways – Indian Highway Autopilot. A lot will depend on the current government and also its policies

Technology is ready, we need the ecosystem to be more prepared now.

Q. Your CEO had a conversation with the PM of India. Could you elaborate on the exchange that happened?

At an event where a lot of young tech CEOs were invited, the PM wanted to know about what these CEOs thought about how the future should look like for India. The meeting was meant for formulating policy. We conveyed that we need to have Autonomous driving trials, having a driver behind the wheel like the ones in California and a couple of other cities. India has the largest number of pedestrians dying in road accidents. The NHAI has come up with a plan to bring that number down by 50%. However, making sure that the infrastructure is correct, that the redlights function and so on and so forth. There is a huge requirement in mandating the Advance Driving Assist System in India. That will not just help but ensure that we are close to that 50% mark by 2020. Now multiple scenarios can make that happen right from the OEMs to the after-market additions level of integrating them. So, if these driver assist systems are made mandatory, there will be ways we can find and implement them.

Q. With the advent of Autonomous technology, have automotive companies started approaching the design and development of their cars differently?

The automotive companies don’t have enough competency to actually come up with autonomous technology that can be useful. Elon Musk and Tesla have set the bar up too high for the other automotive companies. A lot is expected of Automotive companies and they are unable to deliver, for the most part. There has to be a proper system for identifying what the proper quality for autonomous system. For ADAS, it is sorted out. Agencies like EuroNCAP have come up with certain standards that are helping ADAS. Similarly India is also coming up with Bharat NCAP for making sure the guidelines for ADAS adoption. However, for autonomous driving, there is very little by way of standards to measure something against. It is important that everyone comes together to make standards for autonomous driving. According to a RAND study released in 2016, “autonomous cars would have to be driven hundreds of millions of miles and, under some scenarios, hundreds of billions of miles to create enough data to clearly demonstrate their safety” – We can simply say that humans are the best drivers and it is necessary that the autonomous vehicles learn from humans, but it will take a 100 billion miles or more to get there, hence the approach will be incremental where ADAS will lead the way to Autonomy. It is not going to be a sudden shift. It will be a transitional wave and it will be 2025-2030 before we see a marked difference.

Q. There has been quite some talk about initiatives like Open AI. Are these initiatives helping the autonomous movement?

No, actually what these platforms are doing is creating interest. People who love to tinker and play around various machine learning algorithm can now do so and this subsequently helps the AI development. Typically, the AI development was done by specific companies or they would have exclusive rights on it. So, this is kinda democratising it. However, having said that, democratising works well for things like websites, but when it comes to safety critical software systems, OpenAI is creating interest but it cannot be good enough to create deliverables, it can only get people familiar with the concepts and it shouldn’t be an end to and solution for autonomous driving which can be delivered to somebody. So things like Open AI and NVidia are great but they don’t really specify cases where all this can fail and that’s something that adequate standardization can address and these initiatives help us get more people into the ecosystem.

Q. There are companies like Tesla and Google that are developing their own end to end Autonomous driving systems and will certainly come out with their own products. Where does The Hi-Tech Robotics fit in here?

We are technically a B2B company. As far as Tesla is concerned, they are not really competitors. They are more like customers to us. In the products that Tesla makes and brings to the end customers, there are a lot of things that we can do and probably do better and Google is not manufacturing their own cars and is partnering with other OEMs to make cars. So, we aren’t really looking at a business we wouldn’t be working in in the next 2-3 years at least.

QDo you think companies these days that have not worked in autonomous technologies are the ones that are running to companies like Hitech for help to eventually come up with their own products?

There are two aspects. We have two types of customer base – . One who needs help to cope up with sudden change in landscape of Automotive todayand the second ones are those who are aware and know what they want from usand we will compliment them with our tech.. There is some amount of FOMO (Fear of Missing Out) and they are certainly coming to us and we are just providing intelligence to them. We are working with 7-8 OEMs right now and 40-50% hasn’t been working on autonomous tech and the other half have good technology and are still working with us for providing them with our IP to improvise it.


About the author

Team Evo India

Add Comment

Click here to post a comment

Login via any of these platforms to post your comment.