Deploying Autonomous Vehicles in 2025
It’s the year 2025. Alexa alerts you of the time and current temperature outside: 3:40 pm; sunny with light winds from the SE. It sounds like a good day for a drive. You reach for your keys and head to the garage. You get in the car, strap on the seatbelt, and turn the key in the ignition. Off you go…
But something is wrong with the picture.
You are still driving yourself.
Autonomous vehicles are still not commercially available.
The projected time-to-market from 2020 was pushed forward yet again.
Let’s face reality.
It’s the year 2019 and fully autonomous vehicles are far from commercial use. There are countless roadblocks — no pun intended — issues with safety, challenges with training and testing, and in general, a lack of a systems-level approach to problem-solving.
Something big needs to change to deploy autonomous vehicles at scale.
Why the big picture matters?
It’s said that the devil is in the details, and for good measure. Any engineer, developer, or researcher working with autonomous technology knows that a slight misstep or miscalculation in testing can have dire consequences, and where such thinking is natural and beneficial, it is also dangerous and regressive.
Testing is an absolute necessity to make autonomous vehicles function properly and safely but…
Autonomous vehicle training and testing is costly, time-intensive, and requires manual and machine labor.
This raises the question: how does one balance detailed safety requirements with efficient workflows and time investment with the cost of development?
The answer is in a systems-level approach.
What is a System?
“A system is a set of organized and established procedures that governs how an entity functions.”
A system can be small like a car on the road, or large like a city. The principles are the same: a set of rules govern how a car functions internally and how it behaves externally with objects and people around it. Similarly, how a city functions is dictated by intrinsic factors such as geographical location and extrinsic factors such as weather changes.
The overlap in both systems comes from their existence in the physical world, which is an even bigger system. Thus, as smaller subsystems become increasingly dynamic (i.e. smart), the physical world needs to adapt and fit their needs.
Currently, we use all sorts of tools to monitor, control, and optimize how a system interacts with its subsystems but there is still a gap between the system and its subsystems. There is a gap between autonomous vehicles training and testing and our physical world. And even though there is an abundance of tools, they lack efficiency. Such tools:
1. Don’t offer immediate physical solutions,
2. Don’t integrate well within workflows,
3. Reduce functionality to a “point-and-click” interphase.
How do we bridge the gap? The answer is in the words themselves and the inherent nature of smart technology: artificial intelligence.
Artificial Intelligence and Autonomous Vehicles
Artificial intelligence is a core component of the autonomous tech sector. Today, autonomous vehicles are fitted with camera-based machine vision systems, radar-based detection units, driver condition evaluation, and sensor fusion engine control units (ECUs).
Furthermore, deep learning, the most human-like AI, is expected to be the largest and fastest growing technology in the automotive AI market. It forms the basis of systems such as voice recognition, sentiment analysis, motion detection, natural language processing, and recommendation engines.
In 2016, the automotive AI market valued at USD 641 billion and it is anticipated to reach up to USD 10.8 billion by 2025. In fact, the installation rate of AI-based systems in autonomous vehicles is predicted to rise by 109% in 2025. The reasoning is simple:
When technology becomes smarter, the value attached to it also increases; people are more invested in the technology and willing to spend more on it.
Case in point? The rise of smartphones. Today, the technology is a one trick tool for communication and a catch-all platform with constantly evolving functionality.
The progress of the autonomous vehicle industry is primed to rival the evolution of smartphones but the technology needs a proper system, a physical world to thrive in — one that offers immediate, real-time solutions, integrates well with other subsystems, and reduces the need for intermediary “point-and-click” tools.
We need AI-driven physical worlds that allow for training, testing, and validating autonomous vehicles; worlds that provide an end-to-end system that not only bypasses roadblocks (cost, time, and labor) but also becomes an ecosystem that catalyzes growth and accelerates time-to-market.
Now that would be a beautiful world to live in.
