For years, Reuben Brewer had been reading tragic news reports about police officers and motorists being killed and injured during traffic stops gone awry.
The root of the problem, he eventually realized, was fear, the powerful, anticipatory anxiety that can overwhelm a hyper-vigilant officer or cause a motorist to behave in ways that arouse suspicion, leading to a violent encounter.
Brewer knew he couldn't take guns from people who shouldn't wield them or single-handedly remove poorly trained police officers from American streets, but he wondered if there was a way to lower the temperature of everyday encounters between jittery police officers and the large numbers of drivers who fear for their lives each time police lights flash in their rearview mirror.
After 16 months of research and development, Brewer - a senior robotics researcher at the nonprofit SRI International in Menlo Park, California - has unveiled his solution: a robot that allows police officers to conduct traffic stops without leaving the safety of their vehicle.
"The main advantage of a robot over a human is that physical danger no longer matters," Brewer wrote after being reached by email. "The robot is purely defensive, so it can't hurt the motorist. If the motorist damages the robot, it's only money to replace it."
"People are more dangerous when they're scared, so the goal is to remove the possibility of being physically hurt so that they're less scared and less dangerous," he added.
Police officers make more than 50,000 traffic stops each day, according to The Stanford Open Policing Project, an effort that has tracked more than 200 million stops in recent years and found significant disparities in policing.
A survey released by the Bureau of Justice Statistics in 2014 found that black motorists were more likely to be pulled over than white drivers over the previous year. Once stops occur, evidence suggests black drivers are significantly more likely to have their vehicle searched than white drivers.
Traffic stops pose real dangers to officers as well. In 2017, the most recent year of available statistics, 5,108 officers were assaulted during traffic stops and pursuits, an average of about 14 officers each day, according to the FBI's Uniform Crime Reporting.
Philando Castile's death at the hands of a Minnesota police officer in 2016 is perhaps the most high-profile recent example of a police officer killing a motorist who posed no threat. Shot multiple times by Officer Jeronimo Yanez, the 32-year-old's death was streamed on Facebook by his girlfriend. Yanez claimed he thought Castile, who was licensed to carry a firearm, was reaching for his weapon. Castile's girlfriend, Diamond Reynolds, fiercely disputed that claim, but Yanez was acquitted of second-degree manslaughter.
So far, in 2019, 323 people have been shot and killed by police, according to a Washington Post database that has tracked deadly police shootings since 2015.
Amy Shoemaker, a data scientist involved with the Stanford Open Policing Project, told NBC News that a robot is unlikely to completely solve the complex issues that accompany traffic stops.
"If there is a role for this tool, it's to help prevent tragedies like Philando Castile's death," she said, adding that the machine, "doesn't seem like a panacea."
When a vehicle is pulled over, the officer dispatches the GoBetween robot, which is attached to a platform on the driver's side of the police car, by extending a rolling aluminum track to the motorist's window. The robot is perched atop the end of the track, which moves forward on a small wheel. At the same time, a spike strip attached to the robot is lowered to the ground and unfolded between the vehicle's front and rear tires, preventing a potential highway chase. When the encounter is over, the entire contraption retracts and the robot returns to the police car.
With a highway patrol-like helmet perched atop its "head," the weaponless robot includes two video cameras, a microphone and a speaker that allow the motorist and the officer to speak while looking at one another on a screen. A bar-code scanner allows the machine to input driver license information, a signature pad allows the motorist to sign a ticket, and a printer provides the motorist with a hard copy of the citation. Brewer said future prototypes will include a Passive Alcohol Sensor to "sniff for drunk driving."
The robot's senses aren't as sharp as a human officer's, Brewer admits, but he believes the technology can get "pretty close" to approximating a person.
Though the robot may reduce the tension that can spiral out of control during a traffic stop, Brewer pointed out his technology cannot remove human bias from interactions between police officers and motorists.
"Whatever inequalities there currently are with police cars pulling over minorities more often will still be there once there's a robot on their car," he wrote. " he difference is that those interactions (however unequal they may be) shouldn't result in anyone getting hurt or killed."
Brewer said the robot has been shown to four police departments and received mixed reactions. Though some officers embraced the idea, others worried the waterproof robot would be easily broken or give drivers the chance to speed off as the machine is being deployed.
Some officers wanted the robot to be able to peer into other parts of the vehicle to check for weapons or alcohol, a technical change that Brewer said he plans to add. Brewer intends to make future prototypes more rugged, but noted that the current robot withstood eight months of driving attached to a platform on the side of his own car without issue. He plans to move the robot to the front of a police vehicle, where it can be stowed inside the "bull bar," a standard steel barrier on many police vehicles.
Brewer said the next step is testing the machine during real life traffic stops to see how officers -- and the people they pull over -- respond.