When companies first began investing in manufacturing autonomous vehicles, they were pretty optimistic about launching self-driving cars during the 2004 Darpa Challenge. Their hopes were dashed, though, as none of the entrants finished the course to claim that they successfully created a driverless vehicle.

Early adopters soon realized that it would take more than ingenuity, determination, and boundless resources to make self-driving cars a reality. They also needed time, more than a decade, in fact. And while the public may think we are making very slow progress, manufacturers believe otherwise. Experts opine that we’re making great strides since crafting driverless vehicles is comparable to sending people to the moon.

Why Is It Taking So Long to Make Autonomous Vehicles?

As early as 2018, General Motors (GM) promised to provide San Francisco residents self-driving taxis by 2020. It didn’t quite succeed at keeping that pledge despite unveiling a driverless vehicle (one without a steering wheel) in January. The launch has been moved to 2021. What autonomous car challenges are manufacturers facing? We listed down a few below.

Map Creation and Maintenance

Self-driving cars rely on a combination of detailed premade maps and sensors to “see” obstacles on the road in real-time. Before manufacturers can test their creations, they first need to build detailed three-dimensional (3D) maps of every city or town. To do that, employees need to drive the vehicles all over to take pictures. Every image then needs to be categorized into features (intersections, driveways, fire hydrants, etc.).

That is a very time-consuming but necessary process since the maps will ensure that the autonomous vehicles do not get into accidents. In the U.S. alone, that translates to creating and maintaining a database of millions of very detailed street or road images. So, just imagine how much time and effort it would take to create a 3D map of the entire world (since manufacturers are sure to sell driverless cars everywhere), not to mention modifying it to reflect every road’s current state.

Mimicking Complex Social Interactions

Driving is an intensely social process. Drivers need to pay attention to other vehicles and pedestrians. That is easier for humans who rely on common sense than robots who lack that ability.

Self-driving cars, much like any form of artificial intelligence (AI), need “training.” Their software needs to recognize tricky situations that can pop up. They have to consider the hand signals that, say, cyclists use, or accidents may occur. That’s just one example, but as all drivers know, tons of other road challenges can crop up, many of which can be very subtle and unpredictable.

Unpredictable Weather Conditions

Add to those challenges the fact that weather conditions also pose a major challenge. Even humans get into more accidents during bad weather. But as we’ve shown, people can easily adjust their actions, depending on the situation. They don’t need to be trained to slow down when the roads are slippery because of snow; they’ll do so on instinct. Autonomous vehicles, unfortunately, need to be subjected to every kind of weather-related challenge (flooding, strong winds, snow, etc.).

That would take time as well, maybe less than the amount spent on creating detailed 3D maps but still a lot.

Driverless Vehicle Regulations

Before self-driving cars can hit the roads, regulators first need to approve their use. And they will definitely ask about safety. The thing is, we can’t know.

We can never tell when human drivers will get into accidents. In 2019, more than 38,000 people in the U.S. lost their lives to car crashes. In the same way, we can’t tell how many self-driving cars will cause deaths. It will likely take many decades to determine or prove how safe using autonomous vehicles are.

Legal questions such as how to insure driverless vehicles and who is liable (the driver or the manufacturer) in the event of a crash also need answers.

Cybersecurity

Last but not least is the question of cybersecurity. How can manufacturers or drivers ensure their cars can’t get hacked? As we’ve seen with computers and mobile phones, the smarter and more connected they get, the more prone they become to cyber attacks.

Watch this video for a demonstration:

Where Are We Now in the Autonomous Car Space?

Waymo, an autonomous driving tech company in the U.S., is probably the closest to producing a self-driving car. Waymo minivans are already ferrying Lyft passengers in Phoenix, Arizona. Close contenders include Nuro (a robotics company) and Voyage (self-driving car contender). Nuro driverless pods deliver groceries for Kroger in Texas, while Voyage ferries homecare residents from one retirement community to another in Florida.

For now, only these areas seem to have been thoroughly scanned, mapped, and stress-tested. It may take some more years for companies to scale their operations, especially since production requires not just tons of time and effort but also investment. Given the global economy’s state due to the ensuing coronavirus pandemic, we can’t deny that even those with deep pockets are watching their budgets closely.