Indian Startup Built Self-Driving Tech For Indian Roads By Nature-Inspired AI

Think of self-driving cars and the first name that comes to mind is Elon Musk’s Tesla. We’ve all seen the car drive autonomously through American traffic effortlessly.

However, the same cannot be expected for the car in India where traffic rules and lane discipline are seldom followed by busy commuters. No autonomous driving technology can keep up, or so we think, right? Well, not exactly…

You see, a group of engineers from MinusZero in India have developed an autonomous system that promises to work effectively even in Indian traffic conditions. We got to interact with Gagandeep Reehal, co-founder and CEO of MinusZero to better understand how they could bring the autonomous vehicle future to India.

The idea started as an independent research project

The idea started as independent research that was undertaken by the company’s co-founder and CEO, Gagandeep Reehal around ‘AI less dependent on data’ with an idea to solve the problem of costly and extensive data needed for better accuracy when the human brain learns that with minimum data.

Gagandeep explains, “We often give this analogy – ‘You do not show a baby 20,000 images of a bird to make him know there is a bird in front of him. We were working to mimic that human intuition artificially. Self-Driving Vehicle could be one of the most demanding problems that could be solved with tech, considering how much data companies need to collect (e.g., 20 million miles by Waymo, or 3 Billion miles by Tesla) and how costly this part is, and no company is anywhere close to Level 5 autonomy yet. And realizing the immense potential the industry has – autonomous, electric and shared mobility together if executed in the right manner, it was clearly a no brainer.”

Renting a rickshaw to test their autonomous tech

Gagandeep shared that they tested their tech on a rented e-rickshaw since they couldn’t afford a new one and how they worked around it to finally get the tech running, “Since we couldn’t buy an e-rickshaw for modification, we had to rent it out. But we weren’t allowed to open the rickshaw apart since we had to return it in one piece.”

To work around this, they installed control units outside the vehicle, but this came with its own set of challenges, “We took our time to figure that out, and when we tested it in open ground, our software gave a certain steering output, but the handle of the vehicle wouldn’t move. It was stuck. Our motor mechanism couldn’t take the load of the steering. We didn’t have the funds to get a more powerful motor. We tried to solve this for a week, but to no avail. Time was running out.”

He added that this testing was done at a time when a second lockdown was going to be announced across the nation, “Our team members who had come to Jalandhar for on-site work amidst the pandemic had to return back. Another lockdown was coming. We, co-founders, were left alone with a prototype that was not moving at all. We needed to return the rickshaw back in a day. We were devastated, our months of effort by the entire team seemed to go to waste. We had exhausted our savings already.”

However, soon after Gagandeep’s younger brother gave a suggestion that brought them back on track. Gagandeep shared, “Suddenly in the evening, my younger brother made a casual remark – ‘Maybe this rickshaw’s broken, let’s go to the main road and ask some rickshaw-driver to let us see the handle of his rickshaw.’ We went out of sheer curiosity, and we found that normally rickshaw’s handles are pretty smooth. Ours had some defect. We quickly rented the other rickshaw, and one rigorous sleepless night later, we got it working. We did the live test demo the same day. And the very next morning, the lockdown was imposed. That was indeed a tough save by time.”

How MinusZero’s offering stands out

Reehal further explained how autonomous mobility today is eclipsed by three challenges — extensive data need, expensive hardware and dealing with unseen traffic — and how MinusZero has found ways to execute these challenges in an effortless and cost-effective way.

Talking about the challenge with extensive data needs, Rehal said, “One needs to collect millions of miles of data to train the AI neural networks, like the specs I gave in Q2. It’s like if a car learns to drive in the USA, it won’t be able to drive in India. Because it has not been trained on Indian data. This data game consumes years of R&D and millions of dollars in capital. But our proprietary nature-inspired algorithms need only 20000 miles of data to be able to drive anywhere in the world. Our nature-inspired AI is able to extract maximum insights from minimal amounts of data like a human brain does. We do not even need costly high-fidelity maps.”

Talking about costs involved in autonomous vehicle production, he added, “A single LiDAR costs around 9,00,000 INR. We on the other hand eliminate the use of LiDARs entirely relying only on a significantly cheaper camera-focused sensor suite. This happens with our AI that can extract 3D information from a 2D camera image with comparable accuracy.”

For dealing with unseen traffic Reehal tells how their tech differs from what is being used in the industry, “Over we have seen Waymo’s or Tesla’s vehicle crashing on encountering a traffic condition that their AI model has not seen before. So this makes this tech risky for commercial deployment and they are investing time and money to get more and more driving data to train their model on. But our AI mimics that human intuition artificially so it can easily handle any kind of random traffic scenario without any additional data collection needed.”

Developing a solution for Indian roads

Reehal spoke about how India today has a terrible reputation for being impossible for autonomous cars to survive due to the Indian driving conditions, “For one, there is a serious dearth of law-abiding drivers, roads with their dangerously bumpy and unmetalled stretches aren’t the most conducive and basic road infrastructure such as appropriate signages, lane markings and functional traffic lights.”

He added on how their implementation makes it ready for the Indian roads, “In order to counter this, for starters, unlike all of our competitors, we do not rely on detecting lane markings and proper asphalt/cemented roads in order for our system to make driving decisions, instead, our software is designed to map the terrain in front of the vehicle and detect drivable areas, making it, essentially, an all-terrain vehicle. Moving forward, there are various challenges already lined up including lack of regulatory infrastructure for self-driving vehicles, building trust in consumers with regards to safety of the product and some more on similar lines.”

What motivates Gagandeep to keep going

Gagandeep shared how seeing his tech work the way he intended it to, brings him immense satisfaction, “When it comes to building futuristic technologies that come with a great deal of purpose, the biggest satisfaction definitely comes from seeing your technology working as you imagined it to be. Our E-Rickshaw prototype showcased the possibilities that we can achieve with this technology and it provided a great deal of satisfaction seeing the months of hard work coming to reality.”

He added, “Other than that, the entrepreneurial journey in itself, is a grinding yet satisfying journey. As it brings to light the inherent creativity and problem-solving approach that is there inside every individual but doesn’t come out so often due to lack of its requirement. But as an entrepreneur, we are bombarded with different sets of challenges every day but learning ourselves to handle them with perfection, embarks great sense of satisfaction & purpose in us.

---

Source: India Times

Image Courtesy: India Analytics Magazine

---