Robocup @ Home 2022
In the RoboCup @Home, a mobile robot must be able to not only navigate autonomously but also detect, pick up, and place objects in the environment.
Thus we equipped "Cuboid" with a robot arm and hand camera to tackle the challenges above and to discover the potential for wider application.
Information
specification
-
Base : differential drive
-
Max speed : 0.8 [m/s]
-
Dimensions : 370 x 370 x 670 [mm]
-
Weight (base only) : 30 kg
-
Sensors : LIDAR x 5 (PaceCat x3, AkuSense x2 ), HC-SR04 Sonar x6, Kinect V2 x1, MPU-6500 IMU x1
-
Microphone : ANDYCINE AC-M1 (Optional)
-
Microphone array : XK-USB-MIC-UF216 (Optional)
-
PC :
-
CPU : Intel(R) Core(TM) i9-9900 CPU @ 3.10 GHz
-
RAM : 32 GB
-
GPU : GeForce GTX 1650 (4 GB)
-
Tracking of multiple people by integrating information from 2D Lidar and RGBD camera
It is possible to track multiple people by integrating various detection information such as foot detection using 2D Lidar, upper body detection using RGB images, skeleton estimation using Open PifPaf , and 3D skeleton detection using Depth images.
Application example of human tracking system
By applying a human tracking system, it is possible to track the target person while avoiding obstacles.
Object recognition using RGBD camera
Here, it is possible to recognize objects pointed by a person by combining the approximate shape of objects on a plane using point cloud with the 3D skeleton detection mentioned previously. In addition, it is also possible to perform object recognition and 3D position estimation using Deep Learning such as YOLO and YOLACT.
Object grasping without additional information
We have made it possible to extract the approximate shape of an object, estimate graspable points, and generate a grasp plan solely based on the RGBD point cloud information without using any additional information to assist in recognizing the type or shape of the object, such as markers.
6th place prize
We participated in "RoboCup 2022," held in Bangkok, Thailand, and won 6th place.
Competition : RoboCup 2022
Date : July 13-17, 2022
Venue : Thailand (Bangkok)
League : RoboCup @ Home Open Platform League
Team name : Chief Scientist Office (SoftBank Corp., Japan)
