Video Friday: Watch This 3D-Printed Robot Escape (19459000)
RoboSoft 2025 (19459186) : 23-26 Apr 2025, LAUSANNE SWITZERLAND (19659003)
ICUAS 2025 (19459186) : 14-17 may 2025, CHARLOTTE NC
ICRA 2025 (19459186) : 19-23 MAY 2025, ATLANTA GA
London Humanoids Summit: 29-30 May 20,25, LONDON (19659006)
IEEE RCAR 2025: 1-6 June 2025, TOYAMA, JAPAN
2025 Energy Drone & Robotics Summit (19459186) : 16-18 Jun 2025, HOUSTON TX
RSS 2025– 21-25 June 2025 in Los Angeles
ETH Robotics Summer School: 21-27 June 2025, GENEVA
IAS 2025 (19459186) : 30 June-4 Jul 2025, GENOA ITALY (19659011)
ICRES 2025 () : 3-4 Jul 2025, PORTO PORTUGAL (19659012)
IEEE World Haptics: 8-11 July 20,25, SUWON (KOREA)
IFAC Symposium on Robotics (19459186) : 15-18 Jul 2025, PARIS (19659014)
RoboCup 2020– 15-21 July 2025 BAHIA, BRAZIL (19659015)
Enjoy today’s videos.
The robot can walk without any electronics and only by adding a cartridge of compressed gases, straight from the 3D printer. It can be printed all at once, using one material. Researchers from the University of California San Diego, BASF, and describe how they developed this robot in a journal called Advanced Intelligent Systems. They used the most basic technology: a desktop 3-D printer and a commercially available printing material. This design is not only robust but also inexpensive–each robot cost about $20 to produce.
!
[
Paper]via[
University of California San Diego ()
Why would you want a robot that looks like a person to walk? I guess to make it look less weird. But it’s hard for me to imagine how a system with different muscles and joints will be able to move as well as a human.
[
Figure
This little soft robotic worm moves with an incredible level of personality.
The use of soft actuators is essential for the development of soft robots, medical equipment, and haptic devices. Soft actuators are often limited in their applications because they require power to maintain a configuration, and rely on hard-circuitry for control. In this work, the first soft electromagnetic system is demonstrated for externally-controlled bistable actuation or self-regulated astable oscillation.
[
Paper]via[
Georgia Tech
Thank you, Ellen!
If this was a human, a 180-degree rotation of the pelvis would put “break” into “breakdancing”.
This cooking robot impressed my colleagues, but it could be that journalists are always impressed with free food.
[
Posha]
Our latest work is about a hybrid aerial/terrestrial quadruped robot named SPIDAR. It shows unique and complex locomotion in both aerial and terrestial domains, including thrust-assisted creeping motion. This work was presented at the International Symposium of Robotics Research 2024.
[
Paper ]via[
Dragon Lab
Thank you, Moju!
The video, which was captured on Unitree’s testing grounds, shows the rapid progress of humanoid intelligence. Every day is exciting!
[
Unitree
There should be more robots you can ride on.
There should be more robots who wear hats to work.
[
Ugo ]
iRobot is moving away from giant docks that were used for robot vacuums.
[
iRobot ( ]]
It’s a well-known experiment that if you put a fish in a current, it will start swimming just because of the biomechanical design. You can do the exact same thing with a quadruped robot that is not actuated on a treadmill.
[
Delft University of Technology ( )
Mush! Narrowly!
It’s a little scary that this couple seems to be wandering around in a huge mall populated by robots, and no other humans.
[
MagicLab
I’m really trying, but the yellow just doesn’t work for me.
[
Kepler ]
Gap Inc. has reduced injuries and turnover by having Stretch perform the physically demanding task to unload trailers filled to the ceiling with boxes. Employees are also excited about the automation that is meant to keep them safe.
[
Boston Dynamics ]
Ever since arriving on Mars in 2012, NASA???s Curiosity rover ingests samples of Martian soil, rock, and air, to better understand Mars’ past and present habitability. Organic molecules, the building blocks for life, are of particular interest in its search. Curiosity’s onboard laboratory has detected long-chain hydrcarbons in “Cumberland”, the largest organics discovered on Mars.
[
NASA
This University of Toronto Robotics Institute Seminar was presented by Sergey Levine from UC Berkeley on Robotics Foundation Models.
Pretrained general-purpose models have transformed computer vision, natural language processing and other fields. In principle, these approaches are ideal for robotics. Since gathering large amounts data for any robotic platform or application is likely difficult, general pretrained model that provide broad capabilities presents an ideal recipe for enabling robotic learning at scale in real-world applications.
From a general AI perspective, such approaches offer a promising, intriguing approach to some AI challenges. If large-scale training based on embodied experiences can provide diverse physical abilities, this would shed more light on practical questions about designing robots with broad capabilities, as well as the foundations of located problem-solving and physical understanding. Realizing this potential will require overcoming a number challenging obstacles. What data will we use to train robot foundation models? What will the training goal be? How should the alignment or post-training take place? In this talk, I’ll discuss how we can tackle some of these challenges.
The robot can walk without any electronics and only by adding a cartridge of compressed gases, straight from the 3D printer. It can be printed all at once, using one material. Researchers from the University of California San Diego, BASF, and describe how they developed this robot in a journal called Advanced Intelligent Systems. They used the most basic technology: a desktop 3-D printer and a commercially available printing material. This design is not only robust but also inexpensive–each robot cost about $20 to produce.
Paper]via[
University of California San Diego ()
Figure
The use of soft actuators is essential for the development of soft robots, medical equipment, and haptic devices. Soft actuators are often limited in their applications because they require power to maintain a configuration, and rely on hard-circuitry for control. In this work, the first soft electromagnetic system is demonstrated for externally-controlled bistable actuation or self-regulated astable oscillation.
Paper]via[
Georgia Tech
Posha]
Our latest work is about a hybrid aerial/terrestrial quadruped robot named SPIDAR. It shows unique and complex locomotion in both aerial and terrestial domains, including thrust-assisted creeping motion. This work was presented at the International Symposium of Robotics Research 2024.
Paper ]via[
Dragon Lab
The video, which was captured on Unitree’s testing grounds, shows the rapid progress of humanoid intelligence. Every day is exciting!
Unitree
Ugo ]
iRobot ( ]]
Delft University of Technology ( )
MagicLab
Kepler ]
Gap Inc. has reduced injuries and turnover by having Stretch perform the physically demanding task to unload trailers filled to the ceiling with boxes. Employees are also excited about the automation that is meant to keep them safe.
Boston Dynamics ]
Ever since arriving on Mars in 2012, NASA???s Curiosity rover ingests samples of Martian soil, rock, and air, to better understand Mars’ past and present habitability. Organic molecules, the building blocks for life, are of particular interest in its search. Curiosity’s onboard laboratory has detected long-chain hydrcarbons in “Cumberland”, the largest organics discovered on Mars.
NASA
Pretrained general-purpose models have transformed computer vision, natural language processing and other fields. In principle, these approaches are ideal for robotics. Since gathering large amounts data for any robotic platform or application is likely difficult, general pretrained model that provide broad capabilities presents an ideal recipe for enabling robotic learning at scale in real-world applications.
From a general AI perspective, such approaches offer a promising, intriguing approach to some AI challenges. If large-scale training based on embodied experiences can provide diverse physical abilities, this would shed more light on practical questions about designing robots with broad capabilities, as well as the foundations of located problem-solving and physical understanding. Realizing this potential will require overcoming a number challenging obstacles. What data will we use to train robot foundation models? What will the training goal be? How should the alignment or post-training take place? In this talk, I’ll discuss how we can tackle some of these challenges.
