Share the post "It’s raining? AI has the same problems during rainy days as drivers."
Not only drivers have a harder time driving during rainfall. It turns out that artificial intelligence also has problems with driving in such conditions.
The fact that road incidents occur less frequently in the autumn and winter months, which are characterized by an increased amount of rain or snow, results from the fact that we drive slower. Limited visibility, fogged windows, problems with traction and braking on a flooded or snow-covered road make drivers take their feet off the gas. Interestingly, the research on autonomous systems shows that self-driving vehicles also have a problem with the “comfort” of driving in bad weather conditions.
On the one hand, this may seem surprising, but on the other hand, it is symptomatic. Surprisingly, intelligent vehicles are intended to be more perfect and safer, as they will be devoid of human weaknesses. And symptomatic, because every pioneering technology faces certain limitations or barriers, and this is the case with autonomous vehicles and the aforementioned rain, argues Prof. Marcin Ślęzak, director of the Motor Transport Institute.
ITS research carried out under the project “AV-PL-ROAD – Polish road to road transport automation” (in consortium with the Ministry of Infrastructure and the Warsaw University of Technology) indicates a number of factors that currently constitute an obstacle to the dynamic development of self-propelled vehicle technology. Apart from a number of legal issues, organizational and technological shortcomings remain to be solved, such as the construction of the necessary digital infrastructure in the 5G standard, the development of smart cities, data security, electromagnetic interference, calibration of systems for smooth and safe driving at the same time or the mentioned problem of driving in difficult conditions weather.
The artificial intelligence of an autonomous car uses a number of sensors to steer, including radars, LIDARs, ultrasound sensors and cameras. Thanks to them, it detects other nearby objects and makes independent decisions about all aspects of driving. The artificial intelligence system tries to track and predict what other cars will do, whether they are computer-driven or human-driven. This does not mean, however, that in rain or snow, self-propelled vehicles, despite the extensive electronics, will cope better than the proverbial Kowalski. Difficult weather conditions present an additional challenge for intelligent vehicles, as well as for humans. This is because disruptions in the functioning of LIDAR (the most accurate laser sensor onboard a self-driving car), e.g. due to bad weather conditions, are levelled with less accurate radar and cameras. It is not without reason that most road tests of such vehicles are carried out in places where it is sunny most of the time, such as California, Arizona or Texas, notes Prof. Ślęzak.
Of course, this does not mean that the intelligent vehicle will not move in the rain or stop in the middle of the road during heavy snowfall. It will start and keep going, but slower than cars driven by drivers. The factor that will further slow it down will be other road users – people and their conservative and even unpredictable driving. Therefore, it should be expected that self-propelled vehicles (although safer in principle, because they should reduce the number of accidents by about 30%) will generate traffic jams at least in the initial phase of their popularization.
Sources:
– https://bit.ly/3ryt8iM