Pity the poor human driver who must safely navigate a car moving 40-plus meters per second using the same two eyes that we had when we first learned to use fire. Our species has a maximum foot speed of about 10 m/s, so our visual system naturally evolved to keep track of obstacles that are perhaps 20-30 meters ahead of our current position at maximum speed. We might notice a dangerous animal farther out, but a bear or lion that’s 50 meters away is unlikely to pose an imminent threat.
Automobiles augment a human’s natural abilities, allowing us to move much faster than we can travel on foot. When driving, we must now identify and characterize objects hundreds of meters ahead, many of which are also moving at similar speeds and in different directions. With practice, our brains adapt to the driving environment with remarkable success.
Safe driving involves two distinct skill sets. First, we must accurately monitor the constantly changing environment and then manage the car’s motion using our hands and feet to control the direction and speed. Although human drivers can perform these tasks simultaneously, autonomous vehicles (AV) are theoretically capable of doing a better job at both.
It’s no surprise that technology can accurately control the speed and direction of the car. Ralph Teetor patented cruise control in 1950.
1 Before that, Lawrence Sperry demonstrated the first airplane autopilot at the Paris Airplane Safety Competition in 1914 with the pilots actually walking out on the wing to prove the plane was flying by itself.
2 After the Sully incident, American Airlines prefers that pilots and passengers stay off the wings.
Whether you are a human being or an AV, the bigger challenge is identifying and characterizing all the objects in your path. Humans excel at recognizing common objects and know that mailboxes don’t move, other vehicles (should) obey traffic rules and animals are unpredictable. After all, we prove we are not robots by picking out pictures of motorcycles. But our two eyes are no match for the current generation of AVs, which integrate cameras, lidar and radar to sense the dynamic three-dimensional environment.
3 They are certainly more attentive than humans since they are not distracted by the radio, telephone or kids in the back seat, and they don’t get drowsy after a poor night’s sleep.
A 2024 article in Nature Communications studied vehicle accidents involving 35,000 human driven and 2,100 self-driven cars. This data showed human drivers were safer than AV in dawn/dusk conditions. Meanwhile, AV outperformed humans in the rain and are less likely to drive off the road or collide with vehicles and pedestrians at intersections, when merging or stopping.
4
AVs are a quickly evolving technology, and studies like this inform engineers what conditions may make AV algorithms misinterpret sensor information. Data from the early 2020s showed that AVs already had a good safety record and should continue to improve as they gain experience in a real-world driving environment. In the future, AVs will even communicate with each other, forming a network of vehicles that know each other’s thoughts. They will not need to guess why someone’s flashers are on or whether a driver may have forgotten to use their turn signal.
REFERENCES
1.
U.S. Patent 2519859, Speed control device for resisting operation of the accelerator.
2.
www.aerotime.aero/articles/autopilot-flight-automation-history
3.
https://waymo.com/waymo-driver/
4.
Abdel-Aty, M. and Ding, S. (2024), “A matched case-control analysis of autonomous vs human-driven vehicle accidents,”
Nature Communications, 15, 4931,
https://doi.org/10.1038/s41467-024-48526-4.
Tyler Housel is a technologist for Hnuco Technologies and is based in Lansdale, Pa. You can reach him at tylerhousel@comcast.net.