The BBC mentions that Elon Musk of Tesla states that they were very close to having a breakthrough in self-driving vehicle technology. However, if you asked someone in the 50s whether we’d have self-driving cars by the time 20202 rolled around, they’d have probably said yes. Science fiction has sold us the idea of self-driving vehicles for years, and technology has been aiming for this goal. Technology can solve almost any problem that it sets itself to. Unfortunately, with self-driving cars, it’s not just one single problem. In this article, we’ll examine why it’s so hard to realize the dream of self-driving vehicles.
There are Sensors for Everything
If we think about a self-driving car as a person, then for that person to navigate their local area, they’ll need to see and hear everything around them. Sensors are how the autonomous cars “see” and “hear.” The problem is that these sensors are tuned to working in laboratory conditions. In the real world, issues such as weather, temperature, and even graffiti on road signs make it difficult for self-driving cars always to be aware of their surroundings. For self-driving cars to function as expected, all sensors have to work in every possible condition on the planet. It’s a challenge we haven’t quite managed to overcome yet.
Classification of Obstacles
When we see something, we can judge immediately whether it’s passable or impassable. It’s a function we take for granted. For an autonomous car, it’s a lot different. The vehicle will be able to detect obstacles around it, but how does it know it can drive over grass, but not over a cow? Machine learning allows the vehicle to determine what category these obstacles should go in and whether those categories are passable or impassable. Machine learning takes time, however, and it may be several iterations before it can securely classify obstacles in real-time.
Regulations
With the growing use of autonomous vehicles, there’s a push towards legislating their safety standards. Unlike human-driven cars, which have internationally accepted standards for everything from engine wear-and-tear to mobile car battery replacement, new technology is still experimental. There isn’t any regulatory body to define rules because, to date, we aren’t sure what those standards should be. There are rudimentary frameworks, as the Underwriters Laboratory notes can help to define the safety protocols of the vehicle. However, without an internationally accepted standard that is ratified and agreed upon by all stakeholders, self-driving cars can’t be placed on the roads, in the interest of human safety.
No Guarantee for Continued Safety
Labs can set up stringent safety testing, but as vehicles are unleashed into the open road, there’s no control over how they learn. What guarantee does a drive have that his or her self-driving car well be as safe on the open road as it was in the testing lab? There needs to be a system to guarantee consistency across each iteration of the vehicle’s updates. The self-driving vehicle manufacturers have yet to derive a standard that they can agree on to ensure that each successive firmware update or new learning path the vehicle takes will keep its passengers safe.
Accepting the Social Change of Autonomous Vehicles
The most significant problem that self-driving vehicles will run into is the social acceptability of the technology. Self-driving cars will be safer and more efficient than human drivers. However, they are only useful when a similar system runs all other vehicles. Forcing human drivers to hand over the wheel to an AI could be considered an infringement of that person’s rights. Autonomous cars can only become commonplace items when we separate casual, fun driving, from functional, everyday transport.
