STAR1 is a fascinating humanoid robot measuring 5.6 feet tall, weighing 143 pounds. It has a max speed of 8 miles per hour. In the latest experiment, the good folks at Robot Era tested two of these robots in the Gobi Desert in northwestern China. One of them has a pair of sneakers on while the other was bare-footed.
This is the Honey Badger: a quadruped robot from MAB Robotics that can adapt to uneven terrain. It can also walk underwater as this video shows. The Honey Badger 4.0 has 12 DOF of freedom and max payload capacity of 2kg. It can work on grass, concrete, gravel, and other surfaces.
The Honey Badger 4.0 has max speed of 1 m/s. It comes with camera, gas, and LiDAR sensors. It lasts up to 2 hours on battery.
Robosen is behind some of the coolest robots we have tested. This Robosen Buzz Lightyear Interactive Robot is great for kids who happen to be Toy Story fans. It has movable eyes, mouth and lifelike walking motion. It has 23 smart joints for natural movement and supports 34 commands. It also has over 200 audio clips. You can program this robot through an app or web.
This is the Kepler Forerunner K2: a 52DOF humanoid robot for manufacturing, logistics, education, and research. Its software has been upgraded, enabling the robot to autonomously complete tasks in various scenarios. This robot has hands with 15kg load capacity. Its fingertips have flexible sensors with up to 96 contact points. It has various sensors to perceive its surroundings and can walk on complex terrains.
Meet OTSAW’s Camello+: an autonomous multi-purpose delivery robot for indoor and outdoor environments. It can be customized for different users. It has a 360-degree video camera, live audio, ultrasonic sensors, and static/dynamic obstacle avoidance. This robot also comes with a digital touchscreen display.
The Camello+ lasts up to 8 hours on battery. It has a siren alarm and built-in flashing lights for safety. It has omni-wheels and differential steering for rotating on the spot and making tight turns. The above video shows it in action.
[HT]
Unitree’s robots keep going viral on the internet for good reasons. They are flexible, agile, and capable of performing impressive modes. The Unitree G1 is a humanoid robot that can perform standing long jumps up to 1.4 meters. As you can see in the below video, this robot can also stand up after a fall.
This robot can also do full side lunges. This robot weighs around 35kg. It uses a 3D LiDAR and depth camera to perceive its surroundings. The G1 has a starting price of $16k.
[HT]
This is the LimX TRON 1: an impressive multi-modal robot with point-foot, sole, and wheel feet that can take on a variety of tasks. With its soles, it can operate like a legged humanoid robot. Its wheels allow it to adapt to uneven terrain. As this video shows, this robot can climb the stairs, make impressive jumps, and spin pretty fast.
Meet the PIPE-i: a 6-wheel robot designed for challenging conditions. Developed at Beca by Marcus Hall, this 3D scanning robot has a 3D printed polycarbonate/carbon fiber body, 4 LED panels, a LiDAR sensor, a Leica BLK360 3D scanner, and a pan/tilt forward-facing camera. The above video shows it in action:
PIPE-i 6-wheeled #robot for stormwater culvert inspectionspic.twitter.com/fYTXwGmA3Y
— Gadgetify (@Gdgtify) October 17, 2024
PIPE-i can lift itself up to a higher position. It has cameras and proximity sensors. The scanner is inside but flips up when needed to scan.
[HT]
This is the Husky A300: a rugged, customizable mobile robotic platform with 100kg payload capacity, 12hr run time, and 2 m/s max speed that is ready for complex challenges in tough environments. It is IP54 weather resistant. It also has programmable LED lights. Husky ships with ROS 2.
Teaching robots new skills is going to be a lot easier with LLMs. SDS is a see it, do it, and sorted approach that involves quadrupeds learning new skills from a single demonstration. It uses GPT-4o to analyze a video and create reward functions that guide the robot in imitating movements like walking, hopping and running. This system learns through reinforcement learning and improves it performance over time by refining the reward functions.
For this project, a Unitree Go1 robot was used to perform various movements with accuracy. SDS works better than previous methods because it doesn’t need large datasets to train the robot.
[HT] [Authors: Jeffrey Li, Maria Stamatopoulou, and Dimitrios Kanoulas]
