Increasingly, we hear about new developments in robotics as well as the smarts being built into them in order to operate autonomously. We"ve already seen robotic appliances such as the Roomba 980 make the transition from the somewhat random and collision-prone navigation of its predecessors to optically mapping out homes to identify obstacles and execute a round of vacuuming on par with a human. However, dealing with inanimate objects is a different ball game compared to dealing with dozens of moving pedestrians for a robot.
While the task of getting from A to B has been largely solved due to innovations such as GPS and indoor mapping, doing so in a socially acceptable manner to humans adds an additional layer of complexity. Recognizing this challenge, a research team at MIT decided to try their hand at developing a robot that would be up to the task.
What the team ended up building was a miniature rover incorporating an array of "off-the-shelf" sensors including several webcams, a depth sensor, and a lidar sensor, The researchers also leveraged open-source algorithms to tackle the issue of environmental mapping and positioning. Upon that foundation, the group developed Socially Aware Motion Planning with Deep Reinforcement Learning, or SA-CADRL for short, that ensured that the robot kept to the right and passed on the left. Of course, while this is the social norm in the US, the robot could also be trained to adopt left-handed rules depending on the country in which it was operating.
With respect to potential use cases for the robot, lead author Yu Fan "Steven" Chen said that:
"...small robots could operate on sidewalks for package and food delivery. Similarly, personal mobility devices could transport people in large, crowded spaces, such as shopping malls, airports, and hospitals.”
Ironically, should the uptake and integration of robots into our day-to-day lives continue to increase into the future, society could perhaps find itself in a situation where robots end up being collectively more civilized compared to humans when it comes to being good pedestrians. Of course, if MIT"s robot ever got fed up of having to deal with weaving around organic bipeds in hallways, it could dunk itself in the nearest fountain in a manner similar to a security robot that made headlines last month.
Source: MIT News via TechCrunch