Self-driving cars may someday operate on Connecticut roads. New data reveals that Google’s 53 self-driving cars are making progress in California and Texas. However, human passengers in the vehicles have sometimes had to intervene in their operation to prevent car accidents.
Driver intervention helped prevent 13 accidents. During a 14-month period, there were 272 occasions where drivers took control of the cars because software was malfunctioning. There were also 69 occasions when test drivers elected to take control of the autonomous vehicles to ensure their safe operation.
Driver intervention occurred most frequently when the vehicle had a perception discrepancy in which it had problems in visualizing the world. This has been the most difficult problem with developing self-autonomous cars, according to a project developer. A self-driving car may sense that another vehicle turned into its lane while it is actually driving straight. Or it may stop because it senses debris on the road and a human would normally not stop for this debris.
Google said that software discrepancies are the second most common reason for human intervention. The third major reason is unwanted maneuvers, such as unwanted braking or swerving to avoid an obstacle which is not actually a driving hazard.
Experts argue that these findings are not significant and do not reflect on the system’s quality. Google also claims that driver intervention increased in 2015 because the cars are operating in more difficult environments. Technology failures actually decreased since 2014, according to the report.
New technology often has flaws and may never provide absolute safety to vehicle occupants and motorists or prevent an accident caused by a drunk or negligent driver. Fortunately, auto accident victims in Connecticut have the right to seek compensation when another driver’s negligence results in personal injury.
Source: The Washington Post, “Google opens up about when its self-driving cars have nearly crashed,” By Matt McFarland, Jan. 12, 2016