Self-driving cars and safety: key risks
As given by the experts of the automobile industry, safety would become the key advantage of driverless transport. Until now, none of the companies has created an autonomous car safe for both passengers and other traffic participants. The reason lies in the range of unbiased factors. Read on to learn more.
Ignoring a full-line
Lane sensors are embedded in multiple modern cars, including human-driving ones. For example, the BMW 5 Series, Volvo S90, the Mercedes-Benz E-Class, and Tesla. They have been tested by
The the Insurance Institute for Highway Safety (IIHS) to indicate the vehicles’ lane-keeping systems on hills and curves.
The tests showed that all the cars happen to go over the full-line. The BMW 5 Series turned out to be an antihero of the rankings with the biggest number of incidents of going over and touching the full-line. The system disengaged in seven tests out of 14. Tesla Model 3 demonstrated the best results: it touched the line only once when going up the hill.
The obtained results proved the acknowledged truth: driverless cars are far from joining the traffic of public roads.
Misunderstanding traffic signs
Another driverless cars’ problem is the inability to recognize traffic signs if something is painted or pasted on them. IIHS conducted an experiment placing automotive vehicles in front of STOP signs. They were different changes: graffiti or bird droppings.
Even the most unexperienced driver would recognize the sign but automotive cars failed. Their sensors made mistakes: in 66% of cases, a sign with the graffiti was identified as speed limitation STOP. If word STOP was even remotely hidden, they showed wrong results in almost 100% of tests.
Precipitations proved to have a negative influence on driverless cars’ sensors. For example, lidars may take snowflakes for an obstacle. Since snow is falling around the sensors, the vehicle would simply stop moving.
Lasers have demonstrated a bit more effective performance: they draw a 3D map. However, they still make mistakes when recognizing snow hips and icing. A driverless car may take icing for dry asphalt and calculate a braking path, which might lead to an accident.
Therefore, the countries with the snowy weather would see driverless cars not soon.
During the trials of its driverless mode, Tesla Model 3 disengaged 12 times at a 250 km distance. Seven cases were due to the car’s recognition of a tree shadow as an obstacle. Other five incidences happened because the automobile reacted to oncoming vehicles and pedestrians going far ahead.
Under the condition of heavy traffic, such incidents may pose danger to other road users, especially if the driverless car notices an obstacle too late and resorts to emergency breaking.
A funny situation happened in Boston when electric car NuTonomy took a flock of gulls for an obstacle and stopped moving. Pushy birds remained on the road and didn’t react to the noiseless vehicle. The manufacturer thus had to reprogram the autonomous car to keep on moving forward and scare them away.
Inability to work on bridges
Most of driverless cars’ sensors simply do not work on bridges. They literally stop seeing anything. Thus, a human has to take over, which is always risky because he/she may not react timely.
Of course, the idea of making transport fully-automated is great. However, the abovementioned problems are just a mere droplet. Such vehicles cannot be safe unless developers eliminate all the risks.