Testing of the technology continues despite two fatal accidents involving semi-autonomous vehicles occurring within days of each other in March. On April 2, California expanded testing rules to allow remote monitoring instead of having a safety driver in the car. Waymo and another company have since applied to begin testing driverless vehicles in the state. Neighboring Arizona and Nevada also allow testing without a safety driver, but California is the most populous state and is also where many of the companies’ test vehicles are located. States should learn from regulations that foster innovation and safety at the same time.
Setbacks in autonomous vehicle testing
On March 18, an Uber vehicle with a safety driver struck and killed a pedestrian in Tempe, Arizona. Uber immediately halted all testing of self-driving vehicles while investigating the cause of the accident. On March 23, a Tesla vehicle in self-driving mode crashed into a highway median in Mountain View, California, killing the driver. Tesla has not disabled the vehicle’s self-driving capabilities while the company and the National Highway Traffic Safety Administration (NHTSA) investigate the cause of the accident. As advocates tout the improved safety of self-driving cars, these fatalities are likely to invite greater scrutiny of the technology’s claims.
As the name suggests, safety drivers have played a key role in the development of autonomous vehicles. They are specially trained to take over control when the on-board computer encounters a situation where the vehicle cannot navigate on its own. Driving conditions can change rapidly, so safety drivers must remain alert at all times. However, advances in driverless technology are expected to eliminate human input entirely. Without a steering wheel or accelerator pedal, a computer controls the engine and steering based on input from on-board sensors. Passenger shuttles without any of these features have begun operating in Las Vegas, the University of Michigan, and San Ramon, California.
Summary of state laws
While the US Department of Transportation and NHTSA regularly update their guidelines for autonomous vehicles, states have already passed relevant laws. However, there are differences between the two laws on fundamentals, such as the definition of a “vehicle operator.” Tennessee’s SB 151 refers to the Automated Driving System (ADS), while Texas’ SB 2205 specifies a “natural person” in the vehicle. Meanwhile, Georgia’s SB 219 defines an operator as a person who activates an ADS, which may occur remotely in a fleet of vehicles. These differences will impact how states issue licenses for both human drivers and autonomous vehicles going forward.
The most well-known topic is the exemption to following distance rules that allows truck platooning in 11 state laws. Drivers typically maintain an appropriate distance from other vehicles, taking into account speed, road conditions, and human reaction time when traffic stops. Using wireless communication, truck columns can accelerate and brake at a much shorter distance. The shorter distances between vehicles in a truck platoon reduce air resistance for the following vehicles, and fuel savings add up quickly when multiple trucks are hauling cargo over long distances. The popularity of platooning laws suggests a wider focus on commercial applications of autonomous vehicle technology at the state level.
Many other state laws require research into autonomous systems, but none have yet made the results public. Among all states, North Dakota is considering what to do with data generated by autonomous vehicles. SB 2012, enacted in 2017, requires the state Department of Transportation to study “any data or information stored or collected by the use of those vehicles.” A bill rejected that same year, HB 1394, would have given ownership of the data to vehicle owners and allowed them to share the data with customer consent. Given the number of sensors built into autonomous vehicles and the amount of data they generate, determining privacy protections will be a key aspect of any new regulations.
Democracy and self-driving cars lab
The National Conference of State Legislatures’ database of autonomous vehicle legislation tracks the progress of state legislation. Twenty-two states and the District of Columbia have enacted laws, and 10 governors have issued executive orders regarding the operation of autonomous vehicles. Meanwhile, 10 other state legislatures have considered legislation, and eight state legislatures are not considering legislation.
California requires companies that test self-driving cars in the state to report the number of miles driven and the number of times a human driver takes over control from the self-driving system. The number of disengagements per mile driven should drop enough to allow it to relax rules requiring safety drivers.
Within two years, 20 companies had logged more than 1 million combined autonomous miles. Since September 2016, the growth in miles driven has slowed as more states passed laws to attract autonomous vehicle testing. Around the same time, the average number of releases per mile leveled off at 5 per 1,000 vehicle miles, or 1 for every 200 miles driven. This figure remained stable for 14 months, which may have prompted the California Department of Transportation to relax its rules on having a safety driver in the passenger seat.
Two fatal car accidents in March highlight the human cost of testing the technology. If fully realized, replacing human drivers with artificial intelligence could significantly reduce the number of car fatalities, which claimed more than 40,000 lives in the United States in 2016. If the benefits of doing it right outweigh the associated risks, how can engineers and policymakers minimize those risks? National safety guidelines and state laws should incorporate lessons learned from real-world testing. While it may not be possible to prevent all future accidents, they can provide feedback on which policies are most effective and which are not.
Christian Rome Lansang provided research for this blog post.
