Self-Driving Car Accident Statistics

The self-driving car is no longer a futuristic fantasy. The market has become highly competitive, as higher car autonomy levels are likely to come soon. In the United States, autonomous cars have high growth potential in the technological development of automobiles. A research analyst for BI Intelligence gathered data to compile a detailed report that forecasted the exponential growth of vehicle shipment and market penetration.

autonomous-car-shipments-year-over-year

The graph indicates that approximately 320,000 fully autonomous vehicles are expected to be shipped by the end of this year. The number of self-driven vehicles will only continue to increase and penetrate the market in the coming years. Currently, our roads only have partially autonomous vehicles. With the exponential growth of autonomous vehicles, we will eventually have fully autonomous vehicles on the market. The global autonomous vehicle market is expected to reach $37 billion by 2023, as North America has 29% of all self-driving vehicles globally. 

As the technological level of autonomous vehicles expands, there remain pressing concerns about their safety. Over the years, there has been incremental progress in the development of self-driving cars. Some form of driver-assistance technology focused on safety has been installed in most new vehicles. The Society of Automotive Engineers (SAE) defines five levels of driving automation systems; today, vehicles are categorized as level two. At this level, cars control steering, acceleration, and braking, while drivers are still required to remain engaged. Autonomous vehicles are expected to reach level five soon, meaning there will be no need for human interaction. The vehicle will be entirely run by an AI (artificial intelligence) computer system. 

SAE-driving-automation-levels

Autonomous Driving Skepticism

Transitioning into the future is not easy. As technology advances and the fear of the unknown arises, drivers must adjust, and the government will have to keep up. Skepticism about the autonomous market exists. AAA’s (American Automobile Association) annual automatic vehicle survey found that 71% of people are afraid to ride in fully self-driving cars. The survey also found that Americans are more receptive to the idea of autonomous vehicle technology in more limited applications. About half (53 percent) are comfortable with low-speed, short-distance forms of transportation (such as monorails at airports or amusement parks). Approximately 55 percent of Americans believe that by 2029, most cars will have the ability to drive themselves; however, this timeline is overly-optimistic given the number of vehicles already on the road today. Those who are skeptical cite reasons such as lack of trust, not wanting to give up driving completely, and road conditions that won’t support the technology.

Nearly 43% of people in the United States don’t feel safe in autonomous vehicles. Whether they are the driver or the passenger, studies have shown that there is quite a bit of skepticism amongst consumers regarding the safety of autonomous vehicles. A study was conducted from a base of 1,260 consumers from 10 different countries. Their concerns vary as to why they do not trust autonomous vehicles, but they include: 

  • Wanting to be in control at all times 
  • Fear that the car will make mistakes 
  • Driving is a pleasure for them 
  • They don’t know enough about the technology 
  • Don’t trust the cars to manuever through traffic 
  • Unwilling to pay for self-driving functionality 
  • They are concerned that the car could be hacked (cyber attacks)
customer-concerns-on-self-driving-cars

Notably, safety and the ability of autonomous cars to avoid mistakes are among the biggest concerns of people opposed to autonomous vehicles, as a report published by the Boston Consulting Group in 2016 showed. Convincing people of the safety and efficacy of self-driving vehicles will be a challenge for companies as the cars increase in autonomy levels. Although proponents argue that artificial intelligence (AI) will ultimately make driving safer, in 2017, more than 37,000 people in the United States died in traffic-related accidents. Any accident involving an autonomous car must be considered a major setback to a potential driverless future.

Technology is Not Ready Yet

In 2020, an AAA study found that vehicles equipped with active driving assistance systems experience various types of issues every eight miles in real-world driving conditions. Research also shows that active driving assistance systems that combine vehicle acceleration with steering and braking often disengage with little notice, making the driver resume control immediately. This scenario can end in disaster, especially if the driver is momentarily distracted or overly reliant on the system’s capabilities. 

Cyber Attacks 

As vehicles are now advanced beyond their traditional purpose as a means of transport, the requirements for onboard software have exponentially risen. If you are conscious of the automotive industry or the cybersecurity realm, then you would know about some serious cyber security concerns regarding autonomous vehicles. As the automotive industry grows, so does the threat of attack. 

Autonomous, or driverless cars are exceptionally susceptible to cyber attacks as they rely on connectedness to work; using Artificial Intelligence (AI), self-driving vehicles can see the roads and decide how to drive on them. There are six levels of autonomy, ranging from 0 (no autonomy) to 5 (full autonomy); the higher the level, the greater the risk of a cyber attack. The proliferation of technologies embedded in connected and autonomous vehicles (CAVs) has increased the potential of cyber-attacks. The communication systems present between infrastructure and vehicles allow remote attack access for malicious hackers to exploit the system’s vulnerabilities. Autonomous driving combined with increased connectivity poses a substantial threat to the vast socio-economic benefits that are promised by the CAVs. 

The traditional risk assessments are rendered ineffective due to the absence of historical information on cyber attacks. Suppose the car does any substantial computations by connecting to the outside world via the cloud, needs internet connectivity for functionality, or completely relies on outside sensors for decision making. In that case, it might be susceptible to getting hacked.

SAE-automation-levels

Must-Know Statistics on Self Driving Cars 

  • The first autonomous car concept was introduced at New York World’s Fair in 1939
  • Every year, the autonomous vehicles industry increases by 16% globally 
  • Waymo has 600 driverless cars
  • Around 9.1 driverless car crashes occur per million miles driven
  • Over 20 months, Waymo’s self-driving cars have been involved in 18 accidents 
  • In the past four years, eleven Tesla self-driving vehicle accidents have been reported
  • Overall, there have been around 37 Uber test vehicle crashes
  • 55% of small business owners think that in 20 years, their fleets will be fully autonomous

Gathering statistics as developers build autonomous vehicles to their optimal level allows observers to see the change and success over time.

Autonomous Vehicle Crash Rate Comparison

To combat the likelihood of crashes involving autonomous cars in the future, in-depth studies and research are imperative as the innovation of autonomous cars increases to higher levels of autonomy. The general goal of autonomous vehicles is to be at their optimal safety to reach level 5 autonomy. Below are some statistics on autonomous vehicles.

  • According to autonomous vehicle data, the worldwide AV market is presently worth $54 billion.
  • All autonomous vehicles need to include a stop button that can be used in an emergency.
  • Sixteen percent of people would be comfortable allowing a completely autonomous car to drive them about, even if it meant they would have no control.
  • In the United States, 75% of people want Congress to stop self-driving vehicles, indicating that there are still some safety concerns about the technology’s future.
  • According to self-driving car data, 57 percent of consumers indicated they would not feel comfortable buying a self-driving car even if money was not a problem.
  • According to half of US women and two-thirds of men, life and death decisions cannot be taught to any vehicle.
  • In 2021, over 80 firms are currently actively testing more than 1,400 self-driving cars, trucks, and other vehicles in 36 states and Washington, DC.
  • China has the potential to become the greatest market for self-driving cars globally. According to Mckinsey, autonomous cars might account for up to 66 percent of passenger kilometers driven by 2040
  • Overall, autonomous vehicles (AVs) were involved in more crashes: 9.1 crashes per million miles traveled, compared to 4.1 for conventional cars. However, compared to injuries experienced in traditional vehicle collisions, the ones involving injury were minor.
  • In 2015, a self-driving Audi dubbed The Roadrunner completed a 9-day journey from San Francisco to New York. It covered a distance of 3400 miles while passing through 15 states. Despite the presence of a driver, the Audi completed 99 percent of the journey on its own.

Compared to national crash rate estimates of unreported control crashes (4.2 per million miles), the crash rates for Self-Driving Cars operating in autonomous mode when adjusted for crash severity (3.2 per million miles; Level 1 and Level 2 crashes) are lower. These findings reverse an initial assumption that the national crash rate (1.9 per million miles) would be lower than the Self-Driving Car crash rate in autonomous mode (8.7 per million miles) as they do not control for severity of crash or reporting requirements. Additionally, the observed crash rates in the SHRP 2 NDS (Second Strategic Highway Research Program Naturalistic Driving Study), at all levels of severity, were higher than the Self-Driving Car rates. Estimated crash rates from SHRP 2 (age-adjusted) and Self-Driving Car are displayed in Figure 1.

Low exposure for self-driving vehicles (about 1.3 million miles in this study) increases the uncertainty in Self-Driving Car crash rates compared to the SHRP 2 NDS (over 34 million miles) and nearly 3 trillion vehicle miles driven nationally in 2013 (2,965,600,000,000).

crash-rates-per-million-miles-for-autonomous-vehicles

As self-driving cars continue to be tested and companies increase their exposure, the uncertainty in their event rates will greatly decrease. Data suggests that self-driving cars may have low rates of more-severe crashes (Level 1 and Level 2 ) compared to national rates or rates from naturalistic data sets. However, there is still too much uncertainty in self-driving rates to draw this conclusion definitively. However, the data also suggest that less-severe events (i.e., Level 3 crashes) could happen at a significantly lower rate for self-driving cars than in naturalistic settings. 

Interestingly, when the Self-Driving Car events were analyzed using methods developed for SHRP 2, none of the vehicles operating in the autonomous mode were deemed at fault. This is particularly appropriate to vehicles intended for lower-speed use where less-severe events are the most likely to be encountered by the newer generation of the Self-Driving Car fleet.

The First Recorded Pedestrian Fatality Due To Software Malfunction In An Autonomous Car

On May 7, 2018 in Arizona, Elaine Herzberg, age 49, was killed as she attempted to cross a darkened stretch of road with her bike as she was unaware of the vehicle approaching her at thirty-eight miles per hour. It is noted that the vehicle’s radar system registered the pedestrian’s body as she crossed the street only a full six seconds prior to impact, but due to the computer not being programmed to register the common urban phenomenon of a human being walking alongside a bicycle rather than riding it, at first it registered Herzberg as an unknown object, then as a car, then as a bicycle “with varying expectations of future travel path,” the report from the National Transportation Safety Review board stated.  

Rafaela Vasquez, 44, an e-taxi driver, was behind the wheel of the car, a Volvo XC 90, equipped with self-driving technology that Uber was testing almost exactly three months prior on the streets of Tempe, Arizona, where Herzberg was fatally struck. When the car fatally struck Herzberg, it became the first-ever recorded pedestrian fatality caused by an autonomous car in human history. 

An Arizona prosecutor brought a charge of negligent homicide against Vasquez. She had become the first human driver to fail to prevent an imperfect self-driving vehicle from killing a pedestrian. 

A report later revealed that Uber’s technology did not have “the capability to classify an object as a pedestrian unless the object was near a crosswalk.” It was later discovered that the company had disabled the automatic emergency braking functions in the vehicle to reduce the potential for “erratic vehicle behavior.” The company later stated that the car could have been stopped had Vasquez noticed Herzberg and hit the brakes. However, the system was not set up to alert the driver of a possible human being on the road, as Uber had also deactivated the forward-collision warning system.

Records later showed that Vasquez was watching the reality singing competition show The Voice on the streaming service Hulu on her cellphone exactly 12.5 seconds prior to the fatal accident and failed to hit the brakes until after the car had struck Herzberg. Although the driver was held liable, Uber suffered some consequences for its role in the crash. The company settled with Herzberg’s husband and daughter for an undisclosed sum shortly after her death. 

Due to the nature of this accident, discussion of the crash opened the conversation surrounding the ethics of “self-driving” technology and who should be held accountable for car crashes when both a human backup driver and computer system are onboard. An expert stated, “the crash comes down to a vicious cycle: the driver falsely assumed that Uber’s software would be vigilant, and the designers of the software thought the driver would be vigilant.” It can be argued that the companies who develop and deploy these vehicles are indeed the ones driving them — conceptually and morally, if not legally.

Many fear that this incident will only be deemed a matter of distracted driving and may simplify the discussion about a far more complex topic. This incident sparked a lot of questions regarding who would ultimately be at fault for an autonomous car accident. 

The potential for fatal accidents pushes companies and researchers to obtain data on the varying factors that could happen on the road involving an autonomous car. Liability is based on factors regarding negligence of the driver, programming issues, or both. Determining liability can be quite tricky as oftentimes, these cases are multilayered with ambiguity about who was responsible for the accident.

self-driving-car-vision

The image above showcases how the artificial intelligence senses the different variables on the road around the vehicle. For vehicles to drive autonomously, they must perceive their surrounding with the help of sensors, which include camera, ultrasonic, radar, and LiDAR sensors

Capturing the environment accurately is necessary for cars to navigate autonomously in a way that is safe for everyone on the road. Modern vehicles are well equipped with a wide variety of sensors that help them detect their surroundings. This assists drivers with mundane tasks such as parking.

Based on the time-of-flight principle, the radar, cameras, ultrasonic, and LiDAR sensors are key to a vehicle’s sensor’s perception of the environment. 

Cameras have been an integral part of new production vehicles, making maneuvering and parking easier. They also enable systems such as lane departure warnings, blind-spot warnings, and adaptive cruise control. In addition to cameras installed on the vehicle’s exterior, it is planned for them to be inside the vehicle to detect whether the driver is distracted, fnot wearing a seatbelt, or tired. 

Radar sensors (ratio ranging and detection) have gained fame thanks to “radar traps.” In recent years, manufacturers have also installed them in vehicles to measure distances to obtain factual data for systems such as maintaining the space between vehicles and emergency brake assist. Radar technology is directly synonymous with the time-of-flight principle. The sensors emit short pulses in the form of electromagnetic waves, which are then propagated at the speed of light. When waves hit an object, they are reflected and bounce back to the car’s sensor almost instantly.

The shorter the measured time interval between reception and transmission, the closer the object is. The distance of the object can then be calculated based on the speed at which the waves are propagated can be determined with great precision. The sensors can also determine speeds by calculating several measurements. This enables the technology of driver assistance systems such as collision avoidance and adaptive cruise control. 

There is hardly a vehicle nowadays that does not utilize parking assistance. For example, if the vehicle approaches a post, a warning tone will sound, and colored bars will be displayed on the onboard computer. These specific warning signals provide information about where exactly the post is located in the vicinity of the vehicle. This system is made possible by several ultrasonic sensors installed in the bumpers of the vehicle.

Also, based on the time-of-flight principle, ultrasound creates sound waves that are inaudible to the human ear, which are emitted at a frequency of 20,000 Hz. Aside from ultrasonic sensors and parking assistance, ultrasonic sensors are also used to monitor the blind spot for emergency brake assists. 

In contrast to ultrasonic sensors, LiDAR (light detection and ranging) sensors are suitable for long and short range use. Although they have been in existence for many years, they have only been used in vehicles since about the early 2000s. LiDAR is considered a key technology for achieving higher levels of autonomy safely, efficiently, and effectively.