Could A Start-up Entrepreneur Invent A Self-Driving Car In Their Garage?

One of the most popular tropes in high-tech seems to be the notion that an enterprising entrepreneur could miraculously invent the next-gen earth-shattering doodad in their garage.

In fact, the start-up whiz might not even be using their own garage and instead might be toiling away in someone else’s garage, such as one owned by a dear aunt or a neighbor that is okay with having a resident tinkerer that is aiming to change the world.

Time and again, we’ve all heard the stories about the Silicon Valley millionaires and zillionaires that began their arduous journey by using a soldering iron and a cheap keyboard to craft some breakthrough hardware and software for an ingenious system that ultimately took the marketplace by storm.

Here’s a question worth considering: Could a start-up entrepreneur invent a true self-driving car in their garage?

The possibility stokes the imagination.

Perhaps right now there is an engineer or programmer in a nondescript garage that is on the verge of unveiling a true self-driving driverless car.

Can’t wait to see it.

Meanwhile, let’s unpack the matter.

Initial Wild Claims

Those of you that were paying attention to the initial impetus of self-driving cars might remember a few years ago when there were some wild claims by entrepreneurs boldly proclaiming that they could make a self-driving car.

The media loves to hear those kinds of outlandish claims and be the first to tout a potential rags-to-riches tale.

What added to the confusion was the misunderstanding of what constitutes a true self-driving car.

Most would agree that a true self-driving or driverless car is one that drives entirely by itself, using AI technology. There is no requirement for a human driver to undertake the driving for a true self-driving car. The car is being driven by the automation, in an autonomous manner, removing the need for a human to perform the driving task.

We don’t have such a self-driving driverless car as yet.

The roadway trials you hear about are based on a hoped-for driverless car that right now is only able to handle narrow circumstances of driving autonomously. Furthermore, by-and-large those roadway experiments include a back-up or safety driver, a human at the wheel, presumably ready (hopefully) to take over the driving if needed at a moment’s notice.

There are also cars that are considered semi-autonomous that are entering into the marketplace. For a Level 2 and Level 3 car, these semi-autonomous cars require that a human driver be present and must be at the wheel of the vehicle. In that sense, the automation is co-sharing the driving task with the human. These kinds of cars have a type of technology referred to as ADAS (Advanced Driving Assistance Systems).

I am purposely bringing up the distinction between a truly autonomous car and a semi-autonomous car since it makes a big difference in the pursuit of self-driving tech.

In particular, the preponderance of initial claims by eager entrepreneurs was oftentimes mislabeled as an effort to develop a self-driving car, when instead it was primarily a quite narrow effort involving creating a semi-autonomous car (at best).

Unfortunately, the confounding aspect of intertwining true self-driving car possibilities with the piecemeal and less advanced ADAS tinkering led to marketplace bewilderment.

Why were the major automakers being upstaged by some nerdy developer that was working on a shoestring budget and yet apparently was able to eclipse the stuck-in-the-mud backward-looking car manufacturers?

The chagrin and frustration by the automakers could certainly be felt as they had to contend with explaining why their mega million-dollar budgets and hordes of design engineers were unable to do what an entrepreneur was doing in a garage.

Those start-up efforts ultimately failed or faltered, though a few did lead to some acqui-hires (a larger firm buys a company to acquire the talent) or eventually branched out and have continued their quest to this day, usually bankrolled by an automaker or a deep-pockets tech firm.

It Takes A Village

So far, I’ve covered the past and suggested that a garage-based entrepreneurial start-up to develop a true self-driving car was overall a bust.

What about now, would it be possible nowadays?

Unlikely.

Estimates indicate that today’s modern cars have about 150 million lines of code in them, and for a self-driving car it could be somewhere near 250 million lines of code or more.

I assure you that even the most ingenious and fast-paced programmer is not going to create from-scratch the gargantuan 250 million lines of code, it’s beyond any reasonable chance.

That being said, one can argue about the nature of lines of code.

It is well-known that you can potentially do the work of say ten lines of code in one line of code, depending upon which programming language you are using and how you are using it.

Furthermore, some argue that the number of lines of code being quoted for today’s car automation is over-inflated since it includes the code underlying the operating system, the utilities, and other facets of the system that weren’t necessarily written by the self-driving car developers (they instead have licensed or bought the other software to be underneath their software).

You could also point to the gradual emergence of open source software, which is gradually making its way into the self-driving car realm.

With open source, the software source code is made publicly available and anyone can incorporate it into their own system, though bound by certain restrictions and licensing aspects.

Thus, you might occasionally hear the story of a start-up geek that with 2,000 lines of code has made an entire so-called self-driving car, which is usually based on the programmer piecing together the thousands or millions of lines of code from open source repositories and other avenues, and merely writing a small overarching piece that directs or coordinates the rest of the system.

Again, I’m declaring that this is not especially practical, and you shouldn’t be misled by any such claims.

Generally, the rule-of-thumb is that it takes a village to create a self-driving car.

You need teams of experts in a wide range of realms that can come together and synergistically craft a self-driving car. There is a myriad of varying sensors on-board a self-driving car, including cameras, radar, LIDAR, ultra-sonic, and other specialized devices. There are computer processors needed, along with Machine Learning (ML) and Deep Learning (DL) models and capabilities. And so on.

The one-person effort is insufficient.

Consider too the kind of testing needed, encompassing closed track testing, simulation-based testing, and potentially public roadway testing. These are massive efforts. In addition, there is a lot of regulatory paperwork and approvals needed.

If we tried to fit the developers needed into a garage, the garage would have to be the size of a mighty large warehouse.

In any case, the point is that a true self-driving car is so complex, so large-scale, it is not going to be invented by an entrepreneur in a garage.

Lots Of Other Opportunities

For those wide-eyed entrepreneurs out there, I don’t want my assertions herein to crush your dreams.

If you are desirous of doing a startup in the realm of self-driving cars, I want to encourage you to consider doing so. I say this because there are lots of outstanding opportunities in the assorted parts and pieces that go into a self-driving car.

You don’t have to invent an entire self-driving car to become a Rockstar in the self-driving tech industry.

Set your sights on developing the next great Machine Learning algorithm that will vastly improve the ability to detect roadway objects.

Or, craft a new kind of Virtual World model that sits inside the AI system and provides a real-time indication of the traffic situation surrounding the self-driving car.

Maybe develop new hardware such as boosting the existing kinds of sensors on a driverless car, or perhaps invent an entirely new kind of sensor that nobody has yet identified.

The sky is the limit.

Here’s some suggestions of areas that need further help:

·        Sensor hardware and software

·        Data storage and compression

·        Sensor fusion

·        Virtual world models

·        AI system action plans

·        Car controls activation

·        Self-aware AI capabilities

·        Machine Learning

·        Deep Learning

·        Car on-board networking

·        V2V (Vehicle-to-Vehicle)

·        In-car NLP (Natural Language Processing)

·        Distributed AI

·        Car cybersecurity

·        Traffic simulations

·        Car testing tools

·        Etc.

Conclusion

Take a look in your garage, and if there’s someone in there that is pining away at creating a true self-driving car, please gently let them know that it’s going to be a tough road for them.

Probably best to consider tackling the myriad of unsolved problems that still challenge elements of the self-driving car field and see if a solution might be found. Getting your arms around pieces of the puzzle that fit into a true self-driving car provides ample opportunities.

Yes, glory, fame, wealth, and more beckons those of you working in your garages, since you might have the missing piece that completes the quest to develop a true self-driving car.