Back in March, a pedestrian in Arizona was hit and killed by an autonomous Uber, despite having a safety driver behind the wheel. This was the first recorded accident involving a self-driving vehicle with a pedestrian fatality. While Autonomous Vehicles (AV’s) have come a long way, that incident was an important reminder that much more research and development is necessary for AVs before they can become an everyday part of our lives. Overall, these self-driving cars have the opportunity to create a tremendous positive impact in our lives. Understanding more about self-driving cars and how they work can illustrate the potential of their future benefits.
Where it all began
AV’s got their start from a competition held back in 2004, hosted by DARPA, a prominent military research agency. The challenge was to cross a 150-mile course with an unmanned vehicle. The most successful vehicle, Carnegie Mellon’s Sandstorm, traveled a total of 7.4 miles before getting stuck and being unable to free itself. Somewhat disappointing results for such high expectations. However, this didn’t deter DARPA from hosting the competition again. A year and a half later, five of 23 participants were able to finish the entire 132-mile course. This competition was proof that there was a future for AVs.
In 2009, Google set the goal of having a completely autonomous vehicle by 2020. Their self-driving cars have driven more than two million miles since then. In 2016, Google’s AV project took form in “Waymo”. This project paved the way for other companies like Tesla, Uber, and GM to get into the autonomous vehicle game.
How do self-driving vehicles work?
The Society of Automotive Engineers has broken the various categories of automation into five separate levels:
In order to operate at levels 4 and 5, an AV must have the ability to detect its surroundings, predict what will occur after the decision-making process has assessed the situation, and control the vehicle according to its decision.
David Silver, a self-driving car engineer at Udacity, gives an excellent high-level overview of this process in his TED Talk. He breaks down the three most critical functions a vehicle must perform to be autonomous:
Detecting where the car is in relation to its surroundings is determined through a combination of sensors, cameras, radar, ultrasonic sensors, and lidar. Lidar emits pulses of light to create a high-quality 3D map.
While it performs well, it’s bulky and quite costly. Most companies believe lidar is a critical component in a fully autonomous vehicle, but Elon Musk views it as a “crutch.” He’s confident a fully functional AV can be built without it. Detection methods are continually improving as more research is done. For example, Stanford researchers have invented a laser system that allows AVs to see around corners and MIT researchers have developed a sensor that can produce images through dense fog: These sensors work together to determine if there is a solid object in front of the vehicle. Collectively these devices work together to give the car its precise location.
After processing all the information through an array of equipment, the car needs to be able to predict what will happen in the next few seconds. The most common approach is a hybrid of programming rule-based decision making and using a neural network. Using a neural network alone would make it very difficult to understand the process used to reach its decision (especially if it is an error), and using a rule-based system to address every situation would be next to impossible. The hybrid approach relies more on using a neural network powered by data of previous driving situations, with that being supplemented rule-based programming.
Once the car has mapped its ideal path, it must control all the required inputs of steering, braking, and applying the throttle to keep the vehicle traveling safely on the road as it reaches its destination. Below is an image of the Udacity race simulator showing this process, with the yellow line as the ideal path and green line as the vehicle’s actual path.
There are many current challenges that stand in the way of living in a world full of AVs, and it’s likely it will take more than a decade before everything is sorted. The primary bottleneck for development is software. Object analysis and determining how the object will act still needs work. Operating out of ideal environmental situations can throw off this process. Building a decision-making system that reacts like a human is an enormous task to complete. However, both of these issues can be addressed with data and extensive testing to prove long-term reliability.
“ I think we see the merging of several worlds, the tech industry, the internet and the automotive industry. These two worlds merging is like a smart phone on wheels, or you can say it’s a car that has many of the capabilities of smartphones and computers and so on.” — Dieter Zetsche, Head of Mercedes-Benz Cars
As these worlds merge, personal safety is another issue that will need to thoroughly addressed. These AVs will be connected to the IOT, requiring connectivity infrastructure for vehicles to communicate. This also poses a cyber threat, as hackers could steal or intentionally crash a car. Internet security measures would need to be exceptionally secure.
As these systems improve, drivers will pay less attention to the road and be less engaged in the driving process. If a driver isn’t paying attention and gets in an accident that could have been avoided, who is at fault? The driver, or the manufacturer of the car? These vehicles also need a way to carefully bring their passengers to a stop, in case of an emergency.
Currently, there is no federal regulation for autonomous vehicles. Each state has its own set of rules regarding self-driving vehicles. This means a manufacturer has no way of being compliant across the board without regulation revision at a national level.
Another concern is that occupations and transit industries that are built upon our current system may suffer as transportation shifts to AVs. However, AVs will have a positive impact on many other industries, by increasing productivity and reducing costs, thus increasing demand and opening new occupational opportunities.
Despite having many challenges to overcome, self-driving vehicles bring with them an opportunity for an immensely positive impact on our society.
Opportunity for social good
The future of self-driving cars will have an impact that goes far beyond the owners of these autonomous vehicles. A society with readily available access to this type of transportation has an enormous opportunity for social good.
It’s likely the military will have AVs in use before we see mass-use for civilians. These vehicles can safely deliver supplies and provide medical transport to areas of combat. This would offer an increase in safety to soldiers on the battlefield and reduce unnecessary risk.
Once AVs become a mainstream method of transportation, they will make the roads infinitely safer for everyone. 94% of crashes are blamed on human error. Being able to significantly reduce that percentage through the implementation of AVs will save numerous lives and mitigate automotive related injuries. Knowing you’re safe while being transported will allow your travel time to be used in more productive ways.
Overall mobility will increase for everyone. People with disabilities, the elderly, and those too young to drive will all have access to the same, safe and reliable transportation. This has the potential to profoundly increase a person’s quality of life.
Using self-driving cars will help to increase traffic flow and reduce fuel consumption. Benedetto Piccoli of Rutgers University, in conjunction with research from Vanderbilt and Temple University, believe that 40% less fuel will be used. Small accidents that cause big traffic delays can become a problem of the past. With AVs the need for parking diminishes, freeing up valuable space in urban settings for other uses.
Self-driving cars have the potential to make a very positive impact on our society. As these technologies continue to improve and develop, we’ll start to see these benefits bring big opportunities for social good.