Its face is made of a white plastic casing and two cameras. Its black “eyes” focus relentlessly forward, using algorithms and machine learning rather than human reflexes and intuition to spot objects in its path. Two struts pumps the car’s pedals while a metal arm—its three robotic digits attached to the steering wheel—calmly transports human cargo to their destination.
This is IVO, your robot chauffeur.
IVO, short for Intelligent Vehicle Operator, is a different spin on the autonomous car. Usually, self-driving concept cars show no one behind the wheel—if there’s even a wheel at all. That’s is the future fleet of driverless cars we imagine.
But it is unavoidable that the first fully autonomous cars will have to share the road with human drivers and all our many flaws—even after fully driverless cars are common, it’ll be years before humans no longer drive. To ease the transition from the roads of today to the roads of tomorrow, researchers at the Laboratory for Autonomous Robotics at Ben-Gurion University in Israel are developing this suitcase-sized robot driver that turns any car autonomous.
“Right now you have millions of cars that are not autonomous,” Oded Yechiel, one of the electrical engineers developing IVO, told Popular Mechanics. “This in-between phase is kind of complex and needs a solution—this kind of technology will maybe hasten the transition.”
A Robot Solution
When I visited Yechiel at his lab in Israel this past March, he demonstrated the robot for me as it drove a golf cart around the University’s parking lot. The driver was remarkably sensitive to objects in its path and it stopped regularly while easily recognizing obstacles and people. Watching the robot calmly handle this chaotic situation was one a human would handle with much less grace.
Yechiel and his team got the idea for IVO when participating in the 2013 DARPA Robotics Challenge. At that event, the goal was to build a humanoid robot that could walk on difficult terrain, climb into a car, drive it, and then get out and operate various tools.
“What we discovered is we human beings have a long way to go before we’ll be able to build impressive robots like Terminator,” Yechiel says.
Even if sophisticated robo-killers seem unlikely, the DARPA challenge highlighted some promising near-term potential. “We saw driving was straightforward, and this was how the idea was born,” Yechiel says. “We can’t build a million-dollar robot, but we can build the necessities: moving the pedals and the steering wheel.”
After the DARPA challenge, the team spent a year brainstorming and understanding the nuances of robot driving. What are the torques and forces of motion on different steering wheels and pedals? What kind of adaptive control algorithms and neural networks would the robot’s brain need? How many cameras would provide a sufficient view of the world?
The answers to those questions formed the first generation of IVO. The team actually created the software long before building the physical robot over the period of a year and a half.
“The development of software was tested in a physics simulator, which aided to debug many problems before driving a single inch,” Yechiel says.
To get the to the physical robot driver they have today, they “built a roadmap for reaching autonomy.” Starting with a “drive-by-wire” (a semi-automatic driving system), they then integrated object detection and avoidance and built the physical robot.
Now Yechiel’s team are working on the second generation, upgrading the robot with better algorithms, more sophisticated motors, and a “tighter” design. The latest version features four interior cameras (not wanting to complicate installation by requiring exterior cameras), which allows IVO to understand its environment and see objects, vehicles, and pedestrians around it.
With built-in adaptive learning, IVO can adjust and upgrade its driving tactics based on every new vehicle it drives, like when human drivers adjust to a new car’s subtle differences in acceleration and breaking. IVO can learn just as quickly.
Getting Behind the Wheel
Despite all advanced tech required to create a working robot chauffeur, the entire thing is surprisingly portable. To set up IVO, all you need to do is pop the 44-lb. robot into the driver’s seat. It’s two leg rods work the brake and gas pedals and an additional arm does the steering.
“We want the robot to be very compact and when you want to store it you keep it in the trunk and when you want to use it you take it out, unfold it, install it in 5 minutes and then you’re ready to go,” Yechiel says.
While the robot can drive autonomously, you can also give IVO directions with a remote control, like a true backseat driver. With the smartphone app in hand, someone can give IVO instructions using just a few buttons. It accepts directions via GPS coordinates or maps, and can take the vehicle to a series of waypoints or a final destination. The robot does not have a top speed—the maximum speed is set by the vehicle itself, though tests never exceeded 25 mph.
“Driving at high velocities is actually easier since the way is better structured and the driving rules are better defined with less surprises,” Yechiel says. “The robot itself has a navigation sensor that can approximate the velocity, and a decision-making algorithm decides how fast the vehicle should drive…and an adaptive control algorithm decides how much to press the acceleration and braking pedals.”
The robot’s on-board battery lasts about five hours, but can be connected to as many batteries as can fit inside the vehicle to keep it running as long as needed. Ultimately that means the robot is flexible and can adapt to different vehicles. Whether that means driving you down the road or, more likely, taking on a more industrial purpose.
Snow Plows and Dump Trucks
The team is looking for funding and strategic partners to help them commercialize the robot. Yechiel says it’ll still be a year or so until IVO is officially ready to take the wheel, but it might be even longer still before you see one in the fast lane.
At first, Yechiel sees IVO finding a jobs at low-traffic areas or operating work vehicles that move at slow speeds, like airport trucks, tractors, snowplows, street and beach cleaners, and various mining vehicles.
“Complex decision-making and cognition is very complicated,” he says, which makes it more difficult for IVO to overcome regulations to work in populated areas. But on farms, in parking lots, or on mining sites, “there aren’t so many adversaries that are just constantly making your driving task much more complex.”
Yechiel does believe that one day IVO could hit the open road, but until then his team is focusing on developing new technologies for the robot, like giving IVO the ability to hear. By adding acoustic technology IVO, could one day respond to horns, sirens, or human voices.
Maybe even one day, Yechiel says, IVO will be able to “tell bad jokes.”