Gabriel Aguiar Noury
on 9 April 2021
It’s never too late to learn. As any reinforcement learning agent, we get rewarded by the new knowledge that we acquire. Likewise, we learn by doing, by rolling up our sleeves and getting to work. (Do you want a hands-on book on Reinforcement Learning? Here is my personal favourite)
March has shown us great examples of this. From robots learning to encourage social participation to detect serious environmental problems, it was a learning month.
Learning to become more human
In a nutshell, human-robot interaction is a field that studies how to develop robots that are going to work closely with people. This is a fascinating field due to the opportunities it represents. For instance, robots can be used in different emotion recognition therapies with children with autism.
But this study from KTH Royal Institute of Technology illustrates perfectly what robots are able to learn to evoke involvement from people in social contexts. Using a Furhat robot researchers programmed the robot to lead a Swedish word game with participants whose proficiency in the Nordic language was varied. The robot’s face is optical and is created using a high resolution 180 degrees projector, together with face masks.
Researchers found that by redirecting Furhat’s gaze to less proficient players, the robot was capable of eliciting involvement from even the most reluctant participants. This might look like a small study, but it shows how robots can very dynamically influence how people participate, react, and take decisions. Additionally, it contributes to the use of robots in educational settings.
“Robot gaze can modify group dynamics — what role people take in a situation,” Ronald Cumbal, researcher at KTH, says. “Our work builds on that and shows further that even when there is an imbalance in skills required for the activity, the gaze of a robot can still influence how the participants contribute.”
So don’t think that robots are just machines doing repetitive tasks. Given that we want to incorporate robots into our social world, you will likely find more studies that explore robot’s acceptance in different social environments.
Learning to explore the universe
Last month we learned about Perseverance and Ingenuity. But NASA keeps developing new robots to explore Mars. Led by NASA JPL’s Team CoSTAR, they presented the results of the first Martian Analog testing with autonomous quadruped, referred to as Au-Spot.
Perseverance is a wheeled rover. This limits the robot to flat, gently-sloping terrains and agglomerate regolith. Rovers cannot tolerate instability and operate within a low-risk envelope (i.e., low-incline driving to avoid toppling).
Here is where legged robots have an advantage. NASA’s ‘Mars Dog’ is a four-legged robot capable of navigating through hard-to-access planetary surfaces. The robot has unique failure-recovery behaviours, providing a major breakthrough in planetary navigation.
The system comprises a spot robot powered by NASA JPL’s “NeBula” AI package which endows robots with a belief system and higher-levels of autonomy. Spot is equipped with a deep cave exploration payload including an arm.
Mars Dogs operate in synergy, exhibiting collaborative mobility behaviours to accomplish diverse missions that cannot be fulfilled by a single robot. The aim is for these robots to explore the Mars subsurface, where evidence of past life may persist. Ultimately identifying Mars as a potential shelter to future human inhabitants.
Learning to signal environmental disruptions
DraBot is an electronics-free soft-robot, shaped like a dragonfly that uses air pressure, microarchitectures and self-healing hydrogels to watch for changes in pH, temperature and oil.
Developed by engineers at Duke University, DraBot skims across water and reacts to environmental conditions.
You might have heard of soft robots before. They are a growing trend in the industry due to their versatility. For DraBot, the soft principle allows the robot to handle delicate objects, such as biological tissues, that metal or ceramic components would damage. It also helps robots float or squeezes into tight spaces where rigid frames would get stuck. For human interactions, soft robotics is the key in physical intelligent developments increasing the safety of corrobots that physically interact with people.
DraBot is 2.25 inches long with a 1.4-inch wingspan. It was made by pouring silicone into an aluminium mold and baking it. The team used soft lithography to create interior channels and connected them with flexible silicone tubing.
Movement is created by controlling the air pressure in these silicone interior channels. The channels carry air into the front wings, where it escapes through a series of holes pointed directly into the back wings.
- If both back wings are down, the airflow is blocked, and DraBot goes nowhere.
- If both wings are up, the airflow is open, and DraBot goes forward.
The team also designed balloon actuators under each of the back wings close to DraBot’s body. If the balloons are inflated, the wings curl upward. By changing which wings are up or down you are now controlling the direction.
Finally, to detect the pH of water, DraBot uses a self-healing hydrogel painted in one set of wings. Hydrogel is responsive to changes in the surrounding water’s pH. If the water becomes acidic, one side’s front wing fuses with the back wing. This will make the robot spin in a circle, changing its trajectory and signalling researchers of this environmental change.
While DraBot is a proof-of-principle, it could be the precursor to more soft robots that will become environmental sentinels for monitoring a wide range of environmental signs.
Learning to be a vacuum…
Well, it looks like Roomba owners have complained their devices appear “drunk” following a software update.
It seems like the devices were moving in strange directions, constantly recharging or not charging at all.
The devices’ maker iRobot has acknowledged its update had caused problems for “a limited number” of its i7 and s9 Roomba models. Adding that a fix would take “several weeks” to roll out worldwide.
This is another example of why we need to have a plan for updating and rolling back a fleet of robots. Cyberdyne avoided all of these problems in their CLO2 cleaning robots with snaps, ROS and a Brand Store.
Learning from the best
A great engineer, mentor and friend. Someone that started working on Unity 8, moved to snapcraft and snapd, and finished guiding the robotics team. All-terrain, genius coder and a magnificent roboticist.
KUDOS to you and thanks for all.
Here you can find the best robotics tutorials and webinars from one that will always be part of the team:
- From ROS prototype to production on Ubuntu Core – A ROS 5-blog series that takes your ROS prototype to production with snaps and Ubuntu Core.
- Building a commercial robot with open source – Webinar that explores the challenges of commercialising robots and using ROS in your production code.
- Building a production-grade robot on Ubuntu Core – Run-through of the steps needed to build your first production-grade robot from prototype to production.
- Speed up your ROS snap builds – Using stage-snaps to only shipping a single snap to your users or devices.
- How to build a snap using ROS 2 Foxy – Guide to building ROS 2 Foxy snaps using the new ros2-foxy extension.
ROS Kinetic End Of Life
ROS 1 Kinetic is reaching its end of support, together with Ubuntu Xenial. Do you want to know what your options are? Have a look at our latest blogs.
Migrating could be difficult. Maybe you need more time? Or maybe you are concerned about instability of newer releases? Don’t worry. We have something just for you. Learn about ROS ESM, a hardened and long-term supported ROS system for robots and its applications.
Outro
We learn something new every day. Even if we don’t want to 😉 In March, we learned what robotics promises, from social to soft, from updates challenges to planning for the moon.
But we also want to learn from you! Our readers. We’d love to hear about your ROS and or robotics-related project or startup and feature it next month on our blog. Send a summary to
robotics.community@canonical.com, and we’ll be in touch. Thanks for reading.
