In the future, robots might walk alongside people, delivering packages and running errands–but only if they’re able to do it without harming those who get in their way. Mastering the subtle complexities of sidewalk etiquette is a design challenge, as these machines will have to negotiate distracted, slow, and meandering walkers. After all, even some human beings are ignorant to the rules. To tackle this problem, engineers at MIT’s Aerospace Controls Lab are training an autonomous robot to roll alongside people with a technology they call “socially aware navigation.”
There are a few unwritten rules about walking in crowded public spaces that courteous pedestrians abide by: stay to the right, slower foot traffic to the far right, and pass on the left. This helps maintain order and efficiency in a potentially chaotic situation. Unlike robots, which can be programmed to follow a precise path, people tend to exhibit pretty unpredictable behaviors when they walk. They get distracted (ahem, smartphone addicts) and they speed up or slow down at random.
To help their robot perceive this behavior, the engineers hitched a slew of technology to its body, including webcams, a depth sensor, and a high-resolution LIDAR sensor. The robot uses these tools to help it determine where it is, what’s around it, and how best to negotiate obstacles that might get in its way as it travels to its final destination. You can think of this “socially aware” bot as Frogger, but with more finesse.
It was actually interpreting and responding to unpredictable human walkers that posed the real challenge. The engineers used machine learning to understand the social norms that govern the sidewalk. They trained the algorithm through a series of simulations where the robot moved through groups of people until it followed all of the unspoken rules people normally do.
Following these conventional passing rules, the robot’s navigation algorithm calculates the expected paths of people in its way so it can steer around them. Every one-tenth of a second, the robot scans its environment, anticipates the movements of people around it, and reprograms its route accordingly. (To test it, they set the autonomous robot loose in one of MIT’s busiest hallways.)
The engineers see this research as a step toward integrating autonomous robots into our environments and using them for delivery services or even to transport people. “Socially aware navigation is a central capability for mobile robots operating in environments that require frequent interactions with pedestrians,” the study’s lead author Yu Fan “Steven” Chen told MIT News.
With its combination of sensors and artificial intelligence, their robot is more attuned to its surroundings than the average distracted walker (or worse, driver) with their nose glued to their phone. In the future, humans might be better off taking etiquette tips from their local delivery bot.