Google has been testing out their self-driving cars around the company's Mountain Valley, California headquarters for quite some time. Even after all of the testing, the company still thinks humans need to be in the car to take over the controls in case something unpredictable occurs.

Google's self-driving vehicles are able to navigate roads by themselves, but according to recent reports from the California Department of Motor Vehicles, the cars would have hit objects on at least 13 occasions if a human did not take over, CNet reports.

In total, drivers had to manually take control of the self-driving cars 341 times in 14 months. In 272 of those cases, the human had to take over because the self-driving technology failed in the pod-shaped vehicles, according to the report.

The technology failures were due to software errors or perception problems. When these errors occur, the human is given a warning signal from the car to take over the car's controls.

When errors occur during self-driving tests, Google gets a report and the incident is replayed in a simulator. Google found that only 13 of these incidents would have resulted in the car actually crashing without human intervention.

Google is making self-driving cars a priority, and the company hopes to eventually sell software to major car manufacturers. Google believes roads could be safer with self-driving cars, which eliminate the element of human error.

Along with Google, several major car companies like Toyota and Nissan want self-driving cars to be on the roads within the next five years. The testing of these future vehicles has been monitored closely by many interested parties.

Google hopes that by continuing to test these self-driving cars, the incident rate will go down and eventually software errors can be eliminated altogether. In total, the company's self-driving vehicles have logged over 400,000 miles of test driving.

When an incident occurs, Google attempts to issue a "software fix" so that the driverless car can be tested in the same situation, hopefully avoiding the same incident on the next test.