One of the options for the Blog settings is to have “Holiday falling snow” (available until January 4.) I don’t tend to be into that sort of thing, but it did get me thinking about sensors. In my reading (and linking) over the last couple weeks, one of the common concerns was that heavy rain, snow, or other weather conditions might be impossible for automated or autonomous cars to negotiate.
I actually see this as a benefit. The current situation is that a driver determines the safe speed based on experience. There are some locations that do have variable speed limits (description here.) In my personal experience they are seen as information guides not strict limiters. Again based on my personal experience with factory vision systems, vision distance is usually degraded by a poor atmosphere (think snow or rain.) Vision is rarely completely impeded.
Since both machine vision and human vision are impeded by weather, we have a choice. Option one is to let people continue to drive in hazardous conditions, based on their own hazard determinations. Option two is to let autonomous cars determine safe speeds based not only on the vision conditions, but also the ability to control the vehicle based on actual conditions. This determination would factor in many variables, including but not limited to slickness of the road (continuously monitored), vehicle weight (impacted both by type of vehicle, amount of cargo, and number of passengers), visibility, and wind.
The Federal Highway Administration states that up to 23% of accidents have weather implications. That indicates an annual death toll of 6,250. In previous postings, I’ve found that approximately 90-95% of accidents include driver error as a cause. Taking the two together, it seems fairly safe to assume that between 15 and 20% of accidents are related to humans being human, in bad weather. Taking the 80/20 rule — allowing for a drop of 80% of the accidents if autonomous vehicles were driving instead of humans — 5000 fewer roadway deaths can be expected. Let me summarize: annually, 5000 people die because drivers make weather related errors that computers wouldn’t make.
I made a logical jump, so let me explain myself. I suggested that computers would make 20% of the weather related errors that regular drivers make. What mistakes would they catch?
First, they would recognize their degraded visibility. Do you know the safe speed if your visibility is 200 feet instead of 300? Can you even tell what the visibility is? Probably not – but an autonomous car could.
Second, as indicated above, autonomous cars will know when road surfaces are slick, and adjust accordingly. In other words, autonomous cars won’t hydroplane or fishtail. Since multiple sensing technologies can be used the a-car can detect upcoming hazards even if the surface changes suddenly. I’m not going to say that the A-car will see all black ice, but it will see more black ice than I do.
So what happens if the autonomous vehicle’s vision is degraded faster than ours? In other words, what if rain blinds the a-car, but you can still see pretty well? In that case, the a-car should still be able to offer some guidance to drivers who retain primary piloting duties. For example, it may limit speeds because of slick surfaces. Another a-car could signal location. That means that if another car is 500 yards ahead, hidden by rain and fog, your car could provide that guidance.
I started by saying it was a good thing that weather conditions might be impossible for automated or autonomous cars to negotiate. That is because that as driving ability is degraded, the automated car will stay safe. Once the ability falls below a certain level, a person may be able to resume or take control since the autonomous car can no longer operate both autonomously and safely. Even then the car will operate with additional automation safeguards not currently available.