As cars are developed with more digital components inside, a wider array of safety features is now available for consumers. Over the last decade, on-board computers were already taking care of lane-keeping assistance, automatic emergency braking, and various other autonomous functions. Logically, self-driving cars are the next step to automatic vehicle safety.
Let’s start with a definition – an autonomous (or self-driving) car is capable of sensing, analyzing, and adapting to its environment – and able to operate with no human involvement. A fully self-driving car will be able to go anywhere that a traditional car can and perform any function that a traditional driver could.
The Society of Automotive Engineers classifies vehicles into six levels of automation, with level 0 being fully manual and level 5 being fully autonomous. In the U.S. and Europe, Level 3 autonomous cars are the highest classification allowed on the road. However, legal research and testing is being done in preparation for letting Level 4 cars on regular roads. One of the most impressive milestones on the way to Level 4 is happening in the U.K. Level 4 car prototypes will join public roads in three main cities for the course of a year – they will operate on a 9 mile loop selected by Project Endeavor, with driving occurring both day and night.
However, even though we are making huge strides towards Level 4 autonomous car development and integration, we may still be years away from Level 5 vehicles. We’ll give you a brief overview of the hardware and software used in SD cars.
Technology Overview
Different companies have their own vision about which software and hardware should be used to facilitate self-driving capabilities.
Hardware
Let’s take a look at Tesla’s self-driving hardware. They use systems that rely upon cameras and radar. Tesla’s latest version of autopilot hardware (3.0) includes:
- 8 cameras that have a full range of view from all angles;
- 12 sonar sensors that have a 26-foot range;
- Continental radar that has a 558-foot range; and
- 2 custom, Tesla-designed computers.
At the same time, Volvo has their sights set on using Light Detection and Ranging (LiDAR) technology. This high-powered laser sensor is very expensive, with most suppliers selling them for approximately $75,000. However, Volvo has struck a deal with a LiDAR maker, Luminar, which allows the installation of this technology to be more economically feasible.
So, what exactly is LiDAR? Essentially, it is a form of sonar that maps the distance of surrounding objects by analyzing pulsed laser waves. LiDAR facilitates highly accurate depth perception – vehicles that use this technology have a margin of error of only a few centimeters, and the object can be up to 60 meters away. Furthermore, LiDAR works excellently with 3D mapping, so that cars can easily navigate an environment they have visited before.
Software
Hardware provides self-driving cars with input about their surroundings. But, autonomous cars also need powerful software in order to analyze and respond to that input. Here are some capabilities that SD software can give vehicles.
- Perception: SD cars rely upon software to identify and classify objects that the hardware detects. Simultaneously, the direction, speed, and acceleration of these obstacles can be measured and responded to accordingly. For instance, Waymo has perception software that creates a simulated “view” of the environment based upon data collected from radars and sensors. The software then determines whether the car should start driving when a light turns green or, if the lane is blocked, adjust its route.
- Behavior Prediction: SD software draws upon training models to predict the behavior of obstacles in the road. Looking to Waymo again – their software is able to distinguish the difference between cyclists and pedestrians, and accounts for the differences in speed and directional changes.
- Autopilot: Software with autopilot capabilities allow the car to fulfill driving functions under certain conditions. For instance, Tesla’s Autopilot software has auto lane change, autosteer, traffic-aware cruise control, autopark, and car summoning functions.
It’s also important for self-driving cars to have connectivity with online services to receive updates on traffic jams, parking spaces, speed limits, and temporary closures. Check out this video from Huawei to see how their C-V2X software helps cars “see” their surrounding environment.
How Huawei’s C-V2X Cars “See” the World
With so many advances in self-driving software and hardware, why are fully autonomous cars still in the distant future? Put simply, the necessary hardware is incredibly expensive – even with the deal that Volvo made, their LiDaR hardware costs a pretty penny. What’s more, SD software isn’t ready to hit the market yet. There are still plenty of behavior prediction tweaks that must be sorted out, especially in regards to jaywalking – but more on that later.
Where Do We Stand With Self-Driving Cars?
Many corporations want to get in on the RnD phase of self-driving cars – not just automakers. Alibaba, and Google are heavily investing in SD projects. Even the U.S. government is highly interested: the Utah Transit Authority and UDoT are collaborating on a driverless shuttle pilot project: the Easy Mile electric shuttle.
Taking things to a Federal level, the U.S. DoT has publicly declared their commitment to support and facilitate the safe development, testing, and integration of self-driving technologies.
Looking at Texas, we can see a completed pilot project: The Frisco Transportation Management Association held an 8-month pilot program back in 2018. In the program, almost 5,000 riders were transported by self-driving vehicles, with a route extending between Frisco Station, The Star, and Hall Park. After the program was complete, Frisco city engineers collaborated with Texas A&M Transportation Institute to measure consumer feedback on the self-driving vehicles. The results confirmed that further pilot programs are needed throughout the state to help residents place more trust in AI technology in self-driving cars.
Since then, cities around the U.S. and Europe have followed suit. One tiny town, Merrifield, Virginia, recently launched a program, Relay, which is a totally free and 100% electric self-driving shuttle. It will be available to the public for approximately one year, after which public feedback will be measured.
Yet, even though there are plenty of pilot programs around the world, self-driving cars aren’t ready to go public. There are plenty of roadblocks to work through, which we’ll address next.
Current Self-Driving Roadblocks – And Resolutions
While self-driving technology has grown by leaps and bounds, there are still various legal limitations and setbacks. Take, for instance, the death of Elaine Herzberg. In 2018, Uber’s self-driving car struck and killed a woman in Arizona. The reason? The SD software was not designed to recognize pedestrians who weren’t crossing at a crosswalk. When Herzberg jaywalked across the road, it resulted in a crash and her death. After that, Uber suspended its SD test program for months – and, upon returning, the speed of test vehicles was restricted to 25 mph. And recently, the company decided to sell its whole autonomous vehicle research unit to the self-driving startup Aurora.
Legal implications of self-driving cars must be considered. We can look to the American Bar Association for insight on legal questions that must be addressed before Level 4 cars can become available to the general public:
- Who’s at fault? When an SD vehicle crashes into another car, is it the driver’s fault? Is it the AI system’s developer’s fault? Or, perhaps, will the auto manufacturer be held liable?
- How will insurance be handled? When control is given to the vehicle, what criteria classifies risky vs. safe driving (for insurance purposes)?
- Should there be restrictions on what you are allowed to do in the car? Are you allowed to use your phone?
While these questions may be tricky to answer, they are not impossible. Law firms with an expertise in autonomous cars are becoming more common – and, as they set important precedents in court battles, the murky legal implications will become clearer.
What’s more, as SD software continues to develop, cases like Elaine Herzberg’s death will become rare. It has become clear that software developed for self-driving cars must take jaywalkers into account and should be able to make decisions that factor in complex human behavior and ethics.
Benefits of Automation
Even though there are many legal implications to sort through, pilot programs to assess, and RnD to carry out, the benefits of autonomous cars vastly outweigh the investments. In 2019, 38,800 people have been killed on U.S. roadways due to motor vehicle crashes. With the evolution of automated safety technology (of which SD software is a logical step), lives will be saved, and injuries will be prevented.
What’s more, automated vehicles could save up to 50 minutes of commute time per day – thus improving the convenience and efficiency of driving. And that’s not even considering the economic, societal, and mobility benefits that self-driving cars can bring. New vehicle technologies have life-saving and cost-saving capabilities, and it is well-worth investing in their development.
