New product release! Elephant Robotics launches ultraArm, a high-precision desktop robotic arm, with five sets of accessories to help you play the most exciting AI visual games!

In 2020, in order to let more people learn about robotic arms , we launched the world's smallest 6-axis robotic arm: myCobot, followed by the palletizing robotic arm mypalletizer, the small 6-axis mechArm, and the dual-arm myBuddy. We transformed the expensive industrial robotic arm into a desktop-level robotic arm, built a robotic research and education platform at a lower price, and lowered the learning threshold in the field of AI artificial intelligence .

In order to meet the needs of more users, we are constantly expanding our desktop robotic arm product line, launching the ultraArm, a high-precision desktop robotic arm with more powerful functions and easier operation, and pairing it with a complete set of writing, drawing, laser engraving and visual recognition related packages that users can choose directly.

The ultraArm robot arm is an ultra-small desktop robot arm with a classic metal structure design. The main body occupies only half the area of ​​an A4 paper and is flexible to use. Equipped with a high-performance stepper motor, it can achieve ±0.1mm repeatability and high stability. It can be used for writing, drawing, laser engraving, and can be freely matched with accessories such as slide rails, conveyor belts, smart cars, and visual cameras. It can complete different training projects according to different needs and is widely used in higher education, scientific research, teaching, vocational application education and other fields.

Features

Classic structural design

The metallic luster and tough industrial style appearance, 340mm motion radius, and 650g payload help you realize your ideal control scenario.

Compact size

The ultraArm robotic arm takes up very little space, is smaller than an A4 paper, and can be easily placed and integrated into various production environments. It has the characteristics of flexible deployment and rapid response.

High precision

The independently developed high-precision stepper motor can achieve a repeatable positioning accuracy of ±0.1mm.

Quick-change end tooling

Equipped with a quick-change interface, it is compatible with a variety of end effectors. Tools can be replaced with simple installation, greatly improving efficiency.

Long working hours

It can meet the needs of 7 x 24 hours long-term work, so that long-term scientific and technological research is no longer restricted.

Easy to operate

Fast and convenient hardware electrical interface, integrated design, simple operation, laser engraving interface, adaptive gripper interface, pump interface, and burn switch are easily switched. Each student has a machine, so that each student can get a better learning experience through practical exercises.

Graphical programming

MyBlockly is a visual software for drag-and-drop programming. "Drag-and-drop" programming is a basic technique that allows you to drag code building blocks or other visual clues instead of manually writing text-based code. In this way, complex and abstract programming languages ​​become easy to understand.

Mainstream programming languages

Supports Python, C++, Arduino, etc.

ROS2 simulation control

It is developed using ROS, the world's mainstream robot communication framework, and supports simulation, which allows control and algorithm verification in a virtual environment, reducing the requirements for the experimental environment and improving experimental efficiency.

2023

Product Package

As a desktop-level multifunctional open source robotic arm training platform, ultraArm is designed to meet the needs of teaching and practicing robotics technology projects. The practice content involves courses such as robotics mechanism, robot motion control technology, sensor and detection technology, machine vision , robot modeling and simulation, and robot operating system. It can be used for professional course training and professional expansion training for robotics, electromechanical related majors, intelligent manufacturing majors, automation majors, and electronic information majors according to course needs.

ultraArm has 5 packages to choose from, including writing and drawing package, laser engraving package, visual grasping package, slide rail recognition package, and conveyor belt recognition package.

It provides robot scenarios and functional applications, application examples and source code of robot algorithms, allowing users to DIY and learn related knowledge of visual recognition based on their own interests.

Hobby DIY

One-click trajectory generation visualization software, equipped with a work platform, pen clip, laser engraving, warm moments online production. DIY nameplates, mobile phone cases, AI drawing, follow the fashion trends, and enjoy the pleasure of creation.

Set 1: Writing and drawing set

The robot arm can write smooth lines with its high precision and stability. The self-developed Elephant luban software can customize the size of the graphics and the position of the pen. It also provides users with the function of converting pictures and quickly generating running tracks from pictures, giving users more creative space.

Set 2: Laser Engraving Set

If you want to have your own logo, you can also laser engrave it to customize the pattern and logo you want.

Machine Vision

Provides open source AI algorithms, supporting related teaching application scenarios, visual capture, slide rails, conveyor belt sets, static vision, dynamic vision, one-stop experience to quickly get started.

Set 3: Visual Grasping Set

By applying deep learning algorithms, users can use robotic arms to complete positioning, grasping and automatic sorting.

Set 4: Conveyor Belt Set

The distance sensor is used to sense the distance of the logistics to achieve grasping; the logistics are transmitted through the conveyor belt for visual recognition and classification.

Set 5: Slide rail set

Identify objects through vision, track and follow them as they move and grab them.

Using Python programming and the cross-platform computer vision library OpenCV, it has AI capabilities such as target detection, image classification, and intelligent analysis, and supports deep learning and autonomous training.

Machine Vision

Color recognition capture

With the help of high-definition cameras, the ultraArm robotic arm can intelligently sort objects of specific colors through AI vision algorithms.

Image processing uses color threshold binarization based on the LAB color space. Through operations such as corrosion and expansion, the color of the object is extracted and recognized, and then automatically sorted to the corresponding position according to the color.

Image recognition capture

Using the eye-to-hand mode, the camera is used to load the model data trained by Tensorflow through OpenCV to identify image blocks and locate the positions of image blocks. Through the relevant points, the spatial coordinate position of the block relative to the robotic arm is calculated, and a set of relevant actions is set for the robotic arm to put the identified block into the bucket.

Shape recognition grabbing

Object shape recognition is an important direction of pattern recognition. There are many ways to represent the shape of an object in a computer. Based on different shape representations, a variety of shape recognition methods have been proposed, such as Fourier descriptor, principal component analysis, invariant distance, etc. In shape recognition, the pattern features on which the recognition is based are very important.

Through preprocessing: grayscale change of image: grayscale correction, grayscale transformation, histogram correction, image smoothing: neighborhood averaging method, median filtering. Shape extraction (segmentation): edge detection-based method, threshold selection-based method, region generation-based method. Post-extraction processing: shape feature extraction.

QR code recognition and capture

In the feature extraction module, we use the texture feature extraction algorithm to extract multi-resolution histogram features, local binary pattern features, and edge direction histogram features from the original input image. The expression form of these three texture features is one-dimensional array. We connect these three arrays into one one-dimensional array as the input of the subsequent classification module.

When classifying features, we want to keep all image sub-blocks belonging to the 2D barcode and remove all image sub-blocks belonging to the background. In this module, we use the adaptive Spatialboost algorithm.

Feature point recognition and capture

FAST corner detection determines feature points by examining the difference between a pixel and 16 pixels in the surrounding area, and greatly improves detection efficiency through a segmentation test algorithm.

Yolo recognition and crawling

Traditional object detection systems use classifiers to perform detection. To detect objects, these systems use classifiers to evaluate test images at different locations and sizes. For example, object detection systems use deformable parts models (DPM) methods, which use sliding box methods to propose target regions and then use classifiers to achieve recognition. Recent R-CNN-like methods use region proposal methods, which first generate potential bounding boxes and then use classifiers to identify these bounding box regions. Finally, post-processing is used to remove duplicate bounding boxes for optimization. This type of method has complex processes, slow speed and difficulty in training.

We convert the object detection problem into a single regression problem that extracts bounding boxes and class probabilities directly from the image, detecting object categories and locations with just one look (you only look once, YOLO).