Experts from the University of Michigan claim to have found a way to reduce the cost of testing self-driving cars 99.9%
According to experts, consumers will adopt unmanned vehicles, if the test accuracy of 80% will confirm that the robot is 90% safer person. When tested in the usual manner are required for this mileage is about 17 billion km. For reference: over 10 years-hour operation in the city the car can drive just a few million km.
Researchers working at the University of Michigan have found a new way to test self-driving cars, which eliminates the need for huge mileage. This will significantly reduce the cost and accelerate testing.
It is argued that the development based on analysis of over 4 million km real mileage will help to reduce the time required to assess robomobile to cope with potential emergency situation. According to researchers, the reduction of costs reaches of 99.9%.
The main idea of the acceleration lies in the separation of situations into components that can be tested or simulate re-placing robomobile in a kind of concentrate of complex conditions. In this 1000 km test run becomes the equivalent 300 000 — 100 000 000 km in real conditions.
Source: University of Michigan