My robot recognizes the artist of the music captured from the microphone attached to the robot using deep learning. I chose musical compositions by Japanese representative artists suitable for the name of "Arena" as learning data. It was GPU and CPU that did hard work, and I was in a fidget while watching the console output. According to the recognition result of deep learning, the robot changes its behavior a little.
It is a demonstration to steer the robot by stereoscopically viewing the front of the robot using VR goggles for smartphones. Stereo images are created from the two cameras installed in front of the Yamabico robot and transferred to the VR goggle smartphone. In addition, by sending the sensor value of the smartphone to the PC of the robot side, the robot moves according to the movement of the head. We aimed to demonstrate reasonable VR experiences using general devices such as smartphones and Web cameras. Also, by attaching voice guidance and collision avoidance function we tried to achieve user-friendly interface.
I developed a robot which recognizes road signs and changes behavior. Traffic signs has specific colors for drawing driver's attention. So I implemented color extraction and text detection for traffic sign recognition. I used three types of signs: "speed limit", "slow down" and "stop". As for "speed limit", I prepared 40, 70 and 100 km/h signs. When the robot finds the traffic sign, the maximum speed changes according to its contents. In the case of "stop", the robot stops few seconds.
I made the robot that moves by gesture recognition. It uses the laser scanner (URG-04LX) to get the horizontal posision of the hand. When the hand position draws a clockwise circle, the robot rotates clockwise, and when the hand draws counterclockwise circle the robot rotates counterclockwise. When drawing a trajectory that one pushes the robot, it moves forward, and when drawing a trajectory that one pulls the robot, it moves backward.
This is a robot to distribute candy to people raising their hands. After distributing candy, this robot returns to the original position. I implemented this robot by using standard installed sensors in Yamabiko. This robot using OpenPose to recognize people raising their hands. Also, this robot avoiding obstacles using the potential field method based on distance information obtained from range sensor.