The world is eagerly waiting for self-driving cars to become a reality. Before that happens, however, the industry will need to address a number of safety-related issues. There is already a lot of speculation about who would be liable when a self-driving car gets involved in an accident: the driver, the car, or a third party?
This question is a big one for the auto insurance industry, which is worth about $300 billion in the USA. Right now the industry is based on the premise that people drive cars and matters are, for the most part, simple enough to be handled through the legal and financial analysis. However, as self-driving cars become mainstream, the auto insurance sector will go through a major transformation and companies will rely more and more on technical analysis of accidents involving self-driving cars.
Safer self-driving cars are likely to enjoy lower insurance premiums. Since car companies will continue to see their cars subjectively and proclaim them to be the safest, it will be up to the insurance companies to evaluate objective how safe they really are. Insurance premiums can then be offered based on this analysis.
Some companies have already started exploring gritty details of this problem as can be seen in their patent filings. We went through some of these patents to understand how companies are foreseeing to do such accident analysis.
The central question is: how can we determine the extent to which a self-driving car was at fault during a road accident? The methods described in the patents suggest that the insurers might make use of the following data while doing fault analysis.
- What was the point of impact on the car?
- What was the speed of the car and other nearby vehicles prior to the accident?
- Were there any obstructions of the road that were blocking the car’s view?
- Did the car take any protective action to avoid or mitigate the severity of the accident?
- Did any other vehicle (self- or manually-driven) take any actions that caused the accident?
- Were there people walking near the accident site when it happened?
- How were the road and weather conditions?
- Were there any traffic signs or signals (which the car might have missed)?
Clearly, it’s a complex problem and to address it insurance companies will need to analyze the dynamics of a lot of accidents. For this they are also likely to need a database storing past accidents data to serve as a reference.
But where are these companies going to obtain all this data? Most likely – from the cars themselves. Self-driving cars sense, record, and analyze a lot of surrounding data. This data can help the insurance companies to better understand how events unfolded during a particular accident. Then it can be said with more reliability whether or not the car can be blamed for the accident.
A car which does a better job of explaining why it did what it did during an accident is likely to be seen more favorably in these analyses, as opposed to a car that uses a ‘black box’ intelligence system. It is another incentive for autonomous car manufacturers to design the car’s AI in a way that facilitates interpretability of its decisions.
It is also likely that these analyses will also take into account the number of features on which the car relies to ensure safe driving. Some of the important features would be:
- Ability to communicate with other nearby cars to devise a cooperative driving strategy in real time (also sometimes referred to as V2V communication)
- Ability to communicate with roadside infrastructure (V2X communication)
- Accurately simulating likely scenarios to decide a decision path
- Effective communication with human drivers and pedestrians through visual signals to avoid possible confusion
- Ever-ongoing driving training to the car’s AI based on the collective experience gathered by all self-driving cars
Insurance premiums will be lowest for the most technologically advanced cars which make use of all available safety features. This is another push for the car manufacturers to stay on the front lines of innovation.