Pedestrian etiquette – things like not walking into oncoming traffic or keeping to the right of the sidewalk – comes naturally to humans. However, while robots have been programmed to accomplish many things, teaching them to navigate among crowds has proved a challenge because it is hard to accurately predict each person’s path. Now, a team of MIT engineers, led by Steven Chen, have overcome the hurdle with a knee-high autonomous machine that can seamlessly weave itself through people, paving the way for errand-running and pizza delivering robots.
For a robot to be able to move freely among humans, it needs to overcome four major challenges: (1) It must know its place in the world (localization), (2) It should recognize its surroundings (perception), (3) It should be able to identify the right path to the destination (motion planning), and (4) It must follow that path to reach the destination (control).
According to the researchers, addressing the first three problems was relatively easy. To help with localization, they used open-source algorithms to map the robot’s environment and determine its position. The perception issue was solved by outfitting the machine with off-the-shelf products, such as webcams, a depth sensor, and a LIDAR sensor used to make high-resolution maps. To control the robot’s motion they employed technology used to monitor the movement of autonomous ground vehicles.
However, teaching the robot to identify the particular route to follow was not as simple. The project’s co-lead, Michael Everett, says “The part of the field that we thought we needed to innovate on was motion planning. Once you figure out where you are in the world and know how to follow trajectories, which trajectories should you be following?”
Researchers currently take one of two approaches to address the issue. They either use the trajectory-based model, where the robot makes a judgment based on the paths of the people around it, or a reactive model, where it selects the quickest route to avoid collisions. However, both are fraught with problems. The trajectory-based solution takes too long to compute, while the reactive-based model results in the robot either colliding with passersby, or moving around excessively to avoid running into people.
“The knock on robots in real situations is that they might be too cautious or aggressive. People don’t find them to fit into the socially accepted rules, like giving people enough space or driving at acceptable speeds, and they get more in the way than they help,” says Everett.
To overcome the challenge, the team turned to reinforcement learning. This entails using a series of computer simulations to train the robot to move through a crowd by altering its path, based on the speed and movement of other objects in its environment. To make the robot understand the social norms that govern how people behave on sidewalks, they encouraged it to pass on the right, and penalized it when it passed on the left. They also programmed it to scan its environment every one-tenth of a second, and adjust its path if necessary. This allows the robot to move at a constant pace of 3.9 feet (1.2 meters) per second, without pausing to reprogram the route every few seconds.
“We’re not planning an entire path to the goal — it doesn’t make sense to do that anymore, especially if you’re assuming the world is changing. We just look at what we see, choose a velocity, do that for a tenth of a second, then look at the world again, choose another velocity, and go again. This way, we think our robot looks more natural and is anticipating what people are doing.” says Everett.
Chen and his team put their creation to test in the busy, winding halls of MIT’s Stata Building. It worked like a charm, rolling smoothly with the pedestrian flow, generally keeping to the right of hallways, occasionally passing people on the left, and avoiding any collisions. Everett says. “One time there was even a tour group, and it perfectly avoided them.”
While this is a great start, the robot, which was unveiled at the recent IEEE Conference on Intelligent Robots and Systems in Vancouver, Canada is certainly not ready for busy city streets. However, with Chen and his team hard at work, we might soon be learning a thing or two about sidewalk etiquette from a pizza delivery robot!
Resources: fastcodesign.com, MIT.edu
Reading Comprehension (10 questions)
- What does pedestrian etiquette mean?
- Why is pedestrian etiquette hard to teach to robots?
Critical Thinking Challenge
Think on one other problem the robots might run into even if they are...
Vocabulary in Context
“Chen and his team put their creation to test in the busy, winding halls of MIT’s Stata Building.”
In the above sentence, the word winding most likely means: