Join thousands of AI enthusiasts and experts at the Learn AI Community.


Nobody Wants Just a Car: How to Survive the Automotive Competition

Nobody Wants Just a Car: How to Survive the Automotive Competition

Last Updated on April 1, 2021 by Editorial Team

Author(s): Oleksandr Odukha

Self-driving Cars

Things OEMs should consider to increase their chances of staying competitive in the era of self-driving vehicles

According to research by MarketsandMarkets, production of automotive software is poised to grow from a USD 16.9 billion industry in 2020 to a USD 37.7 billion industry by 2025, growing at a CAGR of 16.9%.

The mobility industry is no longer hardware-first with software being an optional frill. Driven by rising demand for advanced safety applications, electrical and electronic components, and connectivity, automotive software development is becoming a focus of automakers. Building a vehicle purely out of metal and mechanical systems isn’t enough to win the competition. People don’t want just cars anymore. They want smart devices on wheels.

To this end, traditional OEMs should act like IT companies. But along with some obvious benefits, this gives rise to new challenges. These include the need to make updates on the fly (throughout the development process and after the product is released), cybersecurity risks and functional safety challenges, compliance complications, the need to process tons of data, and the need to look for more advanced testing techniques.

How can OEMs overcome all these challenges? Here’s what I’ve learned from my extensive experience in engineering autonomous driving software.

Agile and OTA updates

Besides possessing technical expertise, acting like an IT company involves integrating the latest methodologies like Agile into the automotive software development process. Why is this so critical? One reason why automakers struggle to keep up with ever-changing standards, customer needs, and competition is that it takes too long to build a car. Technological progress is much faster. And that’s where Agile comes into play.

The Agile methodology revolves around cross-functional teams consisting of members with different skill sets working toward a common goal. By automating many steps, Agile allows OEMs to focus on meeting business goals, product quality metrics, and security standards.

Agile development and the ability of connected cars to receive over-the-air (OTA) software updates is an invincible combo, making OEMs flexible and adaptable. OTA capabilities not only allow automakers to point out issues before going to market but also allow for continuously improving the product in real time even after its release.

Tesla is a perfect example of how OEMs can successfully leverage the power of Agile and OTA updates. While traditional automakers have to remove entire product lines from the market in case of issues, Tesla acts as a software company, fixing bugs and making updates over the air, in real time.

The moral of this story? Agile and OTA updates are more than just another tech trend. Your customers’ vehicles do need bug fixes on the fly.

Advanced data processing techniques

Another challenge that modern OEMs face is tons of sensor data, which is vital for a seamless driving experience. Computer vision-powered cameras, radar sensors, Lidar sensors, and other sensors can collectively generate up to 25 gigabytes of data per hour. And this number is only increasing. But where does all this data go?

As soon as sensor data is collected, it’s compressed and then processed by artificial intelligence (AI) algorithms. The task of AI is to identify the data required for mission-critical actions and analyze it locally while sending non-critical data to the cloud.

Data is the flesh and blood of a self-driving vehicle, and the higher the level of autonomy, the more data needs to be collected. Since this creates a processing challenge, an expensive machine learning (ML) engine is required. Advanced data fusion and data compression techniques use deep learning technologies to efficiently handle data streams. These techniques will reduce the price of autonomous vehicles, making them available to the mass market.

Built-in security

While data is the flesh and blood of connected cars, it’s also their weak point. Let me unpack this a bit.

The rise of connectivity (which involves collecting tons of data) has led to a drastic increase in cybersecurity threats. Cybercriminals might take advantage of a vehicle’s data security vulnerabilities to steal a driver’s personally identifiable information (including financial information) or even remotely manipulate steering and braking systems. All this puts the driver’s life at risk.

Is the modern automotive industry ready to keep all these threats at bay? Hardly. Since 2017, the number of automotive cybersecurity incidents has increased by 605%. Does that mean automotive cybersecurity is an unachievable target? Of course not. The problem is in the lack of built-in security, achieved by designing systems to work in a hostile environment. To date, this is the only way to build secure systems.

Security and safety must be viewed not as separate elements but as an engineering practice — both from the software and hardware perspectives — and implemented from zero up to the system level. Secure coding standards should be the north star of the development process. Plus, remember that testing shouldn’t be saved for last — rigorous testing is critical throughout the entire product development cycle.

More complications with data compliance

I hate saying this, but I wasn’t completely honest with you in the previous section.

In fact, the most reliable way to protect data is by simply not collecting it. Given that, you should either not collect data at all or collect it according to limitations outlined in the California Consumer Privacy Act (CCPA) and the General Data Protection Regulation (GDPR). Though both of these regimes protect user data, there are certain differences between them.

According to the CCPA, the consumer has the right to know what information is collected about them and how it’s processed, to limit or control the use of this data, to request the deletion of this data, and to opt out of its sale to third parties. The GDPR gives even more control over data processing to the consumer. For example, you can’t even collect a driver’s personal data under the GDPR before they provide explicit consent. Besides, if the user doesn’t consent to data processing that isn’t critical to provide a service, the provider can’t stop offering that service.

Both the CCPA and GDPR have placed a significant burden on OEMs, fleet operators, and car rental providers. They must ensure that systems are in place to let consumers exercise their rights. So how can you ensure 100% compliance? Understand what data you have, assess the risk of data leaks, and consult professionals.

Advanced testing techniques

Though this is the last point, testing shouldn’t be saved for the last.

One of the main reasons why self-driving cars still aren’t on the roads is the inability to test them in every possible scenario. This goes far beyond making sure your car “knows” the driving rules and patterns of behavior in standard situations. What about cases that are hard to predict?

The answer is testing, testing, and again — testing, using advanced testing techniques such as simulated environments and in-the-loop methods. Let’s consider each of these techniques in detail.

A simulated environment models a complete driving scenario, including the driver, traffic, sensor stimulation, and vehicle dynamics. It’s a safe and effective way to validate many aspects of a vehicle as well as to prototype and implement new features.

On the contrary, in-the-loop methods aim to test separate aspects by combining real-world and simulated elements. Depending on the purpose of testing, these methods fall into the following categories:

  • Software-in-the-loop (SIL) tests software in a simulated environment without actual hardware components.
  • Hardware-in-the-loop (HIL) tests hardware elements by putting them into a simulated environment.
  • Vehicle-in-the-loop (VEHIL) is a way to test vehicle performance using elements that simulate other vehicles on the road.
  • Driver-in-the-loop (DIL) is based on putting a real driver into a simulated vehicle that operates in a virtual environment.

All of these methods are actively used to train and test advanced driver assistance systems and have immense potential to revolutionize the self-driving car software development industry, making fully autonomous cars a reality.

Who will help you tackle all these challenges?

Embedded software development for self-driving vehicles calls for competencies that traditional OEMs do not possess at the moment. To make up for it, corporations acquire technology startups, just like General Motors has done. The automotive giant bought Cruise Automation, a self-driving vehicle startup, in 2016 to advance the development of autonomous vehicles and get an edge in the automotive market.

Not only OEMs use this strategy, though. In 2017, the chip manufacturer Intel acquired Mobileye, a leader in computer vision for autonomous driving technology. Mobileye covers a range of technologies including sensor fusion, mapping, crowdsourcing data for high-definition maps, and driving policy intelligence.

This acquisition has brought Intel new relationships with automakers, including with giants like Audi and BMW. It’s been a great step forward for Intel, their consumers, and the automotive industry as a whole.

The second option is to partner with a software development company to develop autonomous vehicle software.

When it comes to outsourcing, I’d recommend looking for the following competencies (in addition to expertise in software development for autonomous driving) in developers:

  • Extensive experience in cloud computing, AI-driven solutions, and AR/VR software
  • Expertise in full-cycle software development
  • Agile mindset
  • Familiarity with ISO 27001/9001, ISO 26262, AUTOSAR 4.0, A-Spice, proprietary navigation data standards, and other international data security standards.

Luckily, there are enough automotive software development companies that meet all these requirements. If you’re looking for top service providers in the industry, you can find a list of them in another article published on Medium.


With the advent of the information technology revolution, software is no longer an additional element of a vehicle. Similar to any other smart device, cars are now commodity hardware for software platforms. That’s why software has become a differentiator of successful automotive projects, creating new challenges associated with data processing, security and safety, data security compliance, testing, and the need to release products and update them rapidly.

No doubt all these requirements are a heavy burden for traditional OEMs to bear alone, making technology partnerships a necessity.

Nobody Wants Just a Car: How to Survive the Automotive Competition was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Published via Towards AI

Feedback ↓