ROSpider robot evaluation report

ROSpider robot evaluation report

Recently, I bought a ROSpider hexapod bionic robot. ROSpider is an intelligent visual hexapod robot developed based on the ROS operating system.

Exterior

In appearance, the ROSpider hexapod robot has six mechanical legs just like its name, and looks like a six-legged spider as a whole. The joints on the legs are bright orange and yellow, and these joints are the key to supporting it to complete various movements. An Astra Pro depth camera is installed above the body of the "spider", which is the eye of the ROSpider robot. It can rotate horizontally and change up and down angles, giving ROSpider a wide range of vision. There are two antennas behind the camera, and the two flexible antennas, like ears, can help the robot receive various signals. The overall feeling of the robot is very smart, full of technological sense and cute at the same time, I am looking forward to its follow-up performance.
insert image description here

ROSpider六足机器人外观实拍图

About the instruction manual

According to the guidance of the development manual for the ROSpider hexapod robot, we installed the ROS system and built the environment. The guidance on this part of the manual is not very detailed, and some problems need to be found and solved on the Internet. Subsequent introductions to various functions and systems, including the structure of the trolley, are very detailed, and the guidance for various experimental operations is also very clear. According to the content of the manual, the test content is divided into: bionic motion, basic AI vision, 3D vision application, deep learning, SLAM mapping and navigation, and far-field microphone array. Next, we will conduct tests according to each section.
insert image description here
insert image description here

ROSpider六足机器人实验手册

Performance Testing

Bionic Motion
After connecting the client and the robot through the robot's own hotspot, and sending simple instructions to the robot through the virtual machine to ensure that it can make corresponding actions, we started to test its bionic motion.
After inputting the relevant instructions, the robot has completed the movement in all directions under the triangular gait and ripple gait forms, and can also adjust the stride and travel speed through instructions; the twisting of the body is also completed very well. It can be seen that the flexibility of the joints is very good, the hexapod cooperation is very coordinated, and the bionic movements are highly completed, which can adapt to more complex terrain environments.
Basic AI vision and 3D vision application
In the robot vision part, we first tested the recognition and tracking functions of the robot's lines and color blocks, and implemented the KCF algorithm with OpenCV vision. Thanks to the high-precision framing measurement of the depth camera, the robot has a high recognition rate of the object to be recognized in this performance. It can be used in scenarios such as line inspection tasks, automated production lines, quality inspection, and material handling. We then tested the configuration of the depth camera to the point cloud image of the environment. The final point cloud image has rich geometric shape and scale information, and is easily affected by changes in light intensity. In most cases, the 3D point cloud information of the surrounding environment can be obtained stably. It has more application prospects in 3D modeling design, automatic driving, medical imaging, etc.
insert image description here

KCF算法实现目标跟随

Screen recording - KCF target follow effect display

Deep learning
In addition, this hexapod robot is equipped with advanced deep learning technology, which enables it to have powerful face recognition and gesture recognition capabilities. Through MediaPipe software, it can quickly and accurately recognize human faces and gestures, so as to achieve a more intelligent interactive experience. In addition, this robot can also perform mission planning and control through the ROS system, making it flexible to adapt to different environments and mission requirements.

Screen Recording - Skeleton Follows

Face recognition
During the face recognition process, the system will return a picture containing the target face and some facial key points. These key points are not too many, but the face can be clearly identified, and the sensitivity and recognition accuracy are very high. This means that even in a complex environment, this hexapod robot can quickly and accurately recognize human faces, thereby improving its interaction ability and intelligence level.
insert image description here

人脸识别截图

Screen Recording - Face Recognition

Gesture recognition
Gesture recognition is to use different gestures as commands to make the robot make different bionic actions. The corresponding action of "pistol" gesture recognition is "attack"; the corresponding action of "bixin" gesture recognition is "twisting body"; the corresponding action of "fist" gesture recognition is "pounce forward". Gesture recognition requires a certain response time, and the recognition accuracy is high. Occasionally, it will correct itself after recognizing deviations.
insert image description here

识别手势并作出对应动作

Screen Recording - Gesture Recognition

SLAM mapping and navigation

In the experimental test of the performance of SLAM mapping and navigation courses, the robot can accurately perceive the external environment, realize SLAM mapping, navigation and path planning, and save the map.
In the experiment guide, it is introduced that the fusion of depth camera vision and lidar is used for map building and path navigation. In the experience of this performance, when the robot was running for the first time, the scanned two-dimensional/three-dimensional image appeared to be deformed to a certain extent. After reducing the traveling speed, it scanned again to obtain a more accurate image.
Therefore, it is recommended that when running this function, the travel speed should be appropriately reduced to ensure the accuracy of mapping. In addition, the robot uses the depth camera data to enhance the recognition rate of special-shaped objects, and obtains ideal results in the obstacle avoidance test.
insert image description here

使用深度相机进行室内建图

General comments

The ROSpider hexapod robot uses Orbi Zhongguang 3D depth camera Astra Plus as the "wisdom eye", and is equipped with high-performance hardware configurations such as Jetson Nano main controller, laser radar, IMU sensor, etc., which can realize map building navigation, path planning, following Obstacle avoidance, voice recognition and other functions, as well as AI vision applications such as object recognition, somatosensory control, target tracking, and mask recognition. With rich functions and good performance, it can meet the needs of many parties, and is suitable for teaching experiments, scientific research and innovative practice in colleges and universities and higher vocational fields.

Guess you like

Origin blog.csdn.net/limingmin2020/article/details/132274499