Autonomous Test Cars Are Coming Soon!

November 28, 2022

Self-driving cars are one of the most exciting technologies on the horizon and one that many different companies are experimenting with right now. They have the potential to save hundreds of thousands of lives by reducing human error in driving and free up significant amounts of time for drivers who no longer have to watch the road. The potential for autonomous cars seems to be limitless. Recently, the federal government released a document that proposed, among other things, to consider allowing testing of vehicles without human inputs like pedals, a wheel, and other parts. In this article, we will discuss self-driving cars and the proposed testing change.

The Rapid Rise Of Automated Vehicles

Self-driving cars connect well with the current age. In the past few years, the most significant change to technology, work, and the economy has been the rapid rise of automation. Everything from factory tasks to white-collar paperwork faces the possibility of automation. Driving is no different. Whether it is long-haul trucks carrying cargo across the country, commuters heading off to work, or ride-hailing services ferrying people to and fro, driving permeates the American experience and automating it would have massive ripple effects. The safety increases would save untold numbers of lives every year.

Cutting Back Road-Related Injuries

Up to ninety percent of all car accidents are a result of human error. Whether that is intoxication, inattention, poor judgment, or poor execution, drivers are responsible for many more accidents than car design flaws or traffic-flow mistakes. That means the fastest and most powerful way to reduce deaths and injuries from driving is to take human error out of the equation. Self-driving cars cannot promise perfect driving, but they may be able to cut down on car-related deaths significantly. An autonomous system that could decrease car deaths by just a half would save hundreds of thousands each year around the world.

A Long Road To Perfection

However, developing autonomous cars is far from easy. It takes immensely complex software and specially-adapted vehicles. The AI running the vehicle needs to learn how to use sensors to identify every kind of obstacle, signal, and combination on the road, integrate them into the local environment, and make decisions. A GPS makes it fairly easy to develop routes for cars to take, but the execution of getting there by obeying traffic laws and driving safely is much harder. In addition to all the development time, autonomous cars need many hours of testing in realistic conditions to improve their accuracy, speed, and flexibility.

Replicating Conditions To Test Autonomous Cars

Even the testing process is laborious. Most car companies that are currently working on autonomous cars have had to build test courses much like the road tests humans face when they obtain their driving license. Managing these courses is expensive and time-consuming; the courses vehicles because they cannot fully replicate the conditions on actual roads, like the behavior of other drivers and unusual road formations. Nevertheless, these courses are a necessary staging ground because they are very low-risk and provide a safe and controlled environment in which to test large, unfinished vehicles.

Test Courses Aren't Enough

The result is that autonomous cars need to have at least some of their testing done on real roads; this is riskier than test courses because it means other drivers or bystanders could be injured if the car makes a mistake. However, the conditions provide much better data and experience for the software and hardware developers working on the car. These vehicles are existing models that have been modified to use the sensors and machinery they need to drive themselves, but all of the usual car elements are intact. This allows a human operator to stay in the driver's seat, ready to take over in the case of a problem.

Removing Human Involvement Entirely

That last point could soon be up for contention. The National Highway Traffic Safety Administration (NHTSA) has raised the idea that they might soon allow testing of autonomous cars that do not have any of the traditional ways for humans to affect the car. Specifically, the NHTSA mentions removing both pedals and the wheel. On the plus side, this would free up valuable space and remove excess weight for something that is ultimately unnecessary if the car functions well. On the downside, human testers no longer have an easy way to retake control of the car should something malfunction.

AIs Could Soon Drive Better Than Humans

Many find the idea of a completely autonomous car to be intimidating, but the reality is that the software and sensors are likely to reach the point where an AI can drive better than a human operator. The difficult part is before then when these vehicles will still require human testing, when the AI still has to share the road with human drivers, and when people still are uncomfortable with having zero control over their car. These concerns are not likely to go away anytime soon. In fact, as car manufacturers clamor for more testing opportunities, they are likely to become more intense.

Machine Learning

Self-driving cars generally use a type of artificial intelligence software called machine learning. Machine learning works by continually improving itself based on its individual performance, learning similarly to a human. That means it can learn quickly, but only when it receives new data and new experiences. There is only so much to learn from a fixed test course, which means there is no way around the need to test autonomous cars on live roads. Moreover, the faster the car can reach a human-free design, the faster it will adapt to optimal driving techniques.

Playing It Safe, For Now

The steering wheel and the pedals are slower and less precise than having the central computer control the wheels and engine directly. An AI that can calculate the physics of road friction, acceleration, and distance will eventually be better at managing a car than any human could be. However, they have not yet reached that point. For now, road tests of autonomous vehicles have kept the speed low to minimize danger and have had a human operator ready to take control at a moment's notice. The number of minor accidents that have occurred is still minimal.

Questioning Legal Liability And Responsibility

No design is perfect, and autonomous cars still have a long way to go, making it virtually certain that autonomous cars will get into more severe accidents during the testing process. Operating a vehicle without human controls opens up many questions about legal liability and responsibility. Current regulations and conventions for insurance and other driving-related concepts will have to radically change as cars become more autonomous. In part, the NHTSA is trying to begin this process of renovating the existing legal structure around cars and driving, but that doesn't make things any more comfortable for those sharing the road with unfinished AIs.

Road-Tests Are Imperative For Progression

There is no getting around the fact that testing will have to take place on live roads and that it will need to be increasingly hands-off so the AI can learn. Perhaps it is better to jump into the process now than hamstring car development by forcing the manufacturers to comply with outdated regulations that soon will lack meaning. There are no easy answers as to when humanity will embrace the transition, and it is likely to be an uneven progression of advances, mistakes, pushbacks, and innovations.

The Road Ahead

Nothing can truly prepare us for fully autonomous cars. But the NHTSA's proposed rules show that the federal agency is thinking ahead. They want to learn from experiences like the drone industry, where the FAA's slow pace of adaptation has begun to push some drone development out of the United States and into less-regulated nations. It will be a long time before the NHTSA implements these rules even if they decide to do so. However, figuring this area out sooner rather than later can expedite the time when self-driving cars can stop being a headline for tech blogs and start becoming a life-saving, accessible invention for all.