Mankaran Singh, a robotics engineer from Bengaluru, wrote this on X around the end of last year: “My alto k10 is being driven by a secondhand Redmi Note 9 Pro that is operating on flowpilot.” Is there anything more Desi than this? In the video, the little hatch is shown traveling in the pouring rain on a highway. The driver is absent.
A 49-second film of FlowPilot, an early prototype of an ostensibly plug-and-play assisted driving system being tested in India, was featured in the viral video. In addition to garnering international notice, it prompted discussion among fans and engineers over the advantages of cost-effective innovation and its hazards.
It seems that Singh, together with his engineering classmates Gunwant Jain and Raghav Prabhakar, have created a prototype of a product that companies like Google, Tesla, and others are investing billions of dollars to develop. FlowPilot is a mobile phone or laptop controlled autonomous driving assistance system. Singh said that about 2,000 volunteers worldwide are trying it.
“We started this project in our third year of college, and the founder of Comma.ai, which started OpenPilot, and George Hotz, who is well-known for reverse engineering the PlayStation 3 and creating iOS jailbreaks, were our main inspirations,” Singh stated.
An open-source advanced driver assistance system (ADAS) is called OpenPilot. ADAS, which is available in more expensive cars in India, allows for various features like driver monitoring, adaptive cruise control, automated lane centering, and lane change aid.
According to Hotz, Comma’s goal is to “solve self-driving cars while delivering shippable intermediaries,” and it’s “looking increasingly like we will win alongside Tesla and Mobileye,” even though ADAS and autonomous operation are not the same thing.
To be sure, Tesla and its competitors operate in nations with higher rates of traffic law compliance than India.
As noted by historian and novelist Adil Jal Darukhanwala, “This technology works brilliantly as a demonstrator in an open field.” It is a formula for disaster in an area where there are many pedestrians, two-wheelers, and people who don’t care about following traffic laws. Our highways are the epitome of chaos.
In addition, Nitin Gadkari, India’s minister of roads, said that autonomous cars will never be used in the country due to the loss of driver jobs.
The creators of FlowPilot, which does not require proprietary gear like OpenPilot, have not been deterred by this.
We made it work with Linux, Windows PCs, and Android phones. All you now need is your cell phone or laptop. You may use it as a radar to assist with driving by downloading certain apps,” he explained.
Cars in severe traffic are shown keeping in their lanes in FlowPilot’s demos on X and Discord.
Tens of thousands of miles have been driven by our roughly 2,000 users. To further hone our machine learning models, we receive the driving data back,” Singh stated.
Apart from the lack of laws governing such vehicles, cars will need to be modified for it to function in India.
“This car is not compatible with phones that you can just plug in and use for ADAS,” Singh stated. “To get this thing working, many adjustments are required. However, roughly 50–60% of cars on the road in the US and Europe would support this straight out of the box. Therefore, it’s lot simpler for those who reside in those nations.
A supported car’s steering system was installed on an Alto K10. Additionally, Singh installed Panda, a universal car interface developed by Comma AI that enables gadgets to communicate with moving cars and give them commands and instructions. The crucial safety code is also kept here.
FlowPilot is ideal for phones.
“The rear camera is used to scan the road ahead, and the front camera is used to monitor the driver.” Everything is there, including a fast CPU, a GPU, and digital signal processing units, according to Singh.
The third level of autonomy in driving assistance, or L3 autonomy, as it is known, is conceivably achievable by cars equipped with the system. This is one level above what premium cars in India offer in terms of ADAS systems.
There is a significant increase in automation from L2 to L3. These autonomous vehicles are capable of making “informed” decisions on their own, including passing a slowly moving car or negotiating a traffic bottleneck. They still need human supervision, though.
Local rivals exist for FlowPilot.
Sanjeev Sharma, an IIT Roorkee graduate, founded the Bhopal-based Swaayatt Robots, which has been developing the technology since 2016. Sharma claims that although his algorithms are more advanced than those of Waymo and Tesla, the business has remained in stealth mode due to a lack of finance and appropriate legislation.
In an X post, Sharma described “Presenting autonomous driving in complex, stochastic and adversarial traffic-dynamics, on the roads in India” and included a video of a Swaayatt autonomous vehicle handling nighttime traffic in Bhopal.
Sharma questioned FlowPilot’s and OpenPilot’s chosen course.
“FlowPilot does behavior cloning. It is a fork of Comma.ai. Behavior cloning is a dead end, as anyone with even a little knowledge of autonomous driving will attest,” he stated. “With this method, you teach the vehicle recovery from every action that is possible by demonstrating to it every scenario that is conceivable, an infinite scenario.”
In order to achieve level 5 autonomy—where no human intervention is required—Swaayatt adopts a more end-to-end approach to the evolving technology, constructing algorithmic models with an eye toward operational cost efficiency and safety.
With 11 data annotators and 15 full-time engineers, the Bhopal-based firm employs Sharma as its chief research scientist. The SUVs in Swaayatt’s fleet are equipped with sensors, cameras, and lidar that are readily available.
Our vehicle of choice is a Mahindra Bolero, which is simple to modify. We installed our own electro-mechanical system—which we designed—in the car to control the accelerator, brakes, and steering. For the upcoming demo, we are getting a Mahindra Thar ready,” Sharma stated.
According to Sharma, Swaayatt is the first firm to effectively demonstrate the ability to drive over uneven terrain and navigate bi-directional traffic on a single-lane road, sometimes requiring the vehicle to move off the road. According to Sharma, major tech businesses are still unfamiliar with this.
In July 2021, Swaayatt secured $1 million in research financing from the Ministry of Electronics and Information Technology (MeitY) in addition to $3 million in seed investment. Before attempting a 100 kmph test on Indian highways, Sharma is currently looking to acquire additional cash.
According to Darukhanwala, there might be a market for these technologies elsewhere.
This is where the technology will be developed. India is a hub for the development of these sophisticated algorithms, but in developed countries where traffic discipline and the rule of law are paramount, their chances of success are higher when applied.
After obtaining authorization from the local police station, Sharma of Swaayatt performs tests during the night. He also drives the car all the time.
“The Motor Vehicle Act underwent its most recent revision in 1988. Since ADAS and autonomous driving are not mentioned, whoever was seated inside your autonomous vehicle will be held accountable if it crashes into someone. I’m the one in charge of all of our exams because of this, and I’m prepared to accept that responsibility,” Sharma stated.
