“Over the river and through the wood,” goes the “New England Boy’s Song About Thanksgiving Day,” based on a poem by Lydia Maria Child, published in 1844. Ever since that time, we’ve associated Thanksgiving with a drive to grandma and grandpa’s house—or at least a road trip to visit family and friends.
Through the years, however, our modes of getting there have certainly changed: the horse-drawn sleigh mentioned in the song is now usually an automobile. And in just a few years, there will most likely be another metamorphosis: from an automobile to a self-driving car.
Within the next decade or two, automated vehicles are expected to be widely available. By 2040, it’s predicted that most cars on the road will be controlled by software—a development that’s expected to save a large number of human lives by replacing easily-distracted, error-prone drivers with high-tech competence.
There’s another positive aspect to this advancement that is less talked about: there’s a potential to save animal lives, too.
Steering away from the soft
On Swedish roads, it’s estimated that there are five accidents every hour involving a wild animal, most commonly deer. The cars are often destroyed, the drivers frequently hurt, and the animals are usually injured or killed.
Here in the United States, roadkill is a huge problem, too. Hundreds of millions of animals die annually in collisions with vehicles. The National Highway Traffic Safety Administration estimates that there are about one million car crashes each year that involve deer alone, killing about 200 people in the process.
There are some measures already in place to make streets and highways safer for wildlife, including roadside fences and specialized crossings. Yet these are frequently expensive and only somewhat effective. The hope is that fully automated vehicles will make accidents with animals a thing of the past.
The first step toward reducing collisions with wildlife may be driving more slowly, which autonomous cars could do. Equipping the vehicles with computerized animal-warning systems would be the second.
Several automated car manufacturers are currently developing systems that use infrared cameras or radar to detect large animals near roads. But these technologies only scratch the surface of what’s possible: driverless cars in the future could use a combination of cameras, radar and lasers to navigate roadways and identify objects in or near their paths. Large animals will be relatively easy for autonomous vehicles to dodge. Since manufacturers make avoiding pedestrian collisions a top, nonnegotiable priority, any animal that resembles a pedestrian will benefit under the same umbrella of caution.
To perceive and protect smaller animals, the systems could be fine-tuned. Internet-linked navigation systems could follow variable speed limits that drop when animals are most active or when cars approach collision hot spots. Some, for example, might provide information about an object’s surface—whether it is hard and likely made of glass, metal or steel; or soft and presumably composed of flesh and fur. Any large, soft object will be treated like a pedestrian.
Prognosticating the parameters
Because of their sometimes-erratic behavior, however, wild—as well as domesticated—animals could prove to be tough for intelligent vehicles to outsmart. Nature is imperfect and unpredictable, and it isn’t clear yet how the rigid calculations of computers will handle that reality.
For example, spotting a moose on the move is one thing. Forecasting its next steps, however, is probably beyond the current state of the art. As of now, we can only make rough predictions based on an animal’s current position and speed. So when an animal is standing still, it’s assumed it will keep standing still until movement is detected.
But there certainly are some things technicians could drop into an algorithm, such as time of day, time of year, a road’s environment and width, and the amount of traffic on it; parameters that affect the likelihood of encountering an animal on the road. Then, the cars could refer to their “cost maps”—systems that tell an automated vehicle at any given moment what objects are currently in the vicinity and how costly it would be to collide with them. A pedestrian would probably be associated with the highest costs, as would a semitruck or any other large vehicle. Hitting a big animal could be linked to the same risk.
Searching for scientists: conservation and citizen
For this potential to be realized, however, conservation biologists will need to get involved: collecting data on roadkill distribution, studying relationships between speed and collisions, and generally engaging in vehicle design now rather than later. One of these innovations already exists: there are citizen science tools for mapping roadkill, and the data could be incorporated into GPS navigation systems.
Without such guidance from conservationists, automated vehicle manufacturers will likely be concerned only with collisions that risk human lives. Creatures too small to cause more than a bump, such as squirrels, won’t be deemed important. Ironically, it’s even possible that animal-vehicle collisions could become more frequent if automated cars are permitted to operate at higher-than-usual speeds, if the industry doesn’t consider their effect on wildlife. Consumers will need to pressure these production companies to consider animal as well as human safety.
Driverless cars could be good for all of us, human and nonhuman alike. According to the National Highway Traffic Safety Administration, more than 94 percent of crashes involve human-driver-related errors. But driverless cars don’t get drunk. They don’t text and drive. And they can be programmed not to speed.
After all, it’s not just people who use the roads.
Here’s to finding your true places and natural habitats,