Neural Network-powered 3D Printable Spider Robot

After nine months of hard work, Roboeve is ready to bring to their neural network-powered “Xpider” to the world.

Xpider Roboeve

In the last couple of years, deep learning techniques have resulted in robots becoming better than humans at tasks such as facial and object recognition; even Google open sourced its machine learning and deep neural network research tool, TensorFlow, in an effort to push machine learning forward. Now, with the help of Intel, a startup from China has developed a 3D printable robot designed to put a form of this powerful AI in the palm of your hand.

Xpider is the smallest commercially available programmable spider robot in the world, measuring only 3.4-inch in diameter. Roboeve, the group behind the project, includes members of Team Water, one of top college robotics teams in China, who previously won three RoboCup championships; one of the top robotics competitions in the world.

To ensure the best quality possible for Xpider production, Roboeve has visited factories all over China, deciding to partner up with Padmate in Xiamen, who have been manufacturers for Lenovo, Nokia, ZTE and other major brands.

Inspired by the Tachikoma from Ghost In The Shell, Roboeve began working on the Xpider in early 2015, aiming to develop an accessible desktop robot with the power of their championship winning competition robots.

Roboeve has also already released the 3D design files for anyone to download and modify under a CC BY-NC-SA license on Wevolver, providing almost limitless customisation options for Xpider owners, while supporting the maker movement’s collective efforts to democratise technology.

With the Xpider’s integrated camera, the user can see exactly what the Xpider sees via the smartphone or tablet app, and thanks to its SmartNode software, anyone can easily program the Xpider using a computer or iPad with Roboeve’s simple drag and drop system.

The Xpider isn’t limited to drag-and-drop programming either, the $129 version can also be trained using a neural network inside its Intel Curie module. Facial recognition, distance sensors, and a remote control can be used to teach the Xpider how to react to people, obstacles, sounds, lights, and more.

The button-sized Intel Curie includes a six-axis combo accelerometer and gyroscope sensor, which powers the robot’s movement tracking and gesture recognition. The Curie also includes an Intel Quark SE microcontroller with an advanced pattern-matching engine to enable the robot to “learn”.

Pretty amazing for something that weighs only 150 grams.


Please enter your comment!
Please enter your name here