6.1. Overview

In this chapter, an environment is provided that combines object recognition using yolox_ros and grasp pose estimation using graspnet_ros , and performs Pick & Place with hsrb_interface using the estimated grasp pose.

This environment consists of two Docker environments: yolox_ws and hsrb_pnp_ws . Their respective roles are described below.

6.1.1. yolox_ws

• Using yolox_ros , the grasping target object is detected from an image sequence.

• By using compressedImage and compressedDepth as input, data transmission is accelerated.

• The detection results from yolox_ros are fed into graspnet_ros , which estimates and outputs the grasp pose.

6.1.2. hsrb_pnp_ws

• Based on the estimated grasp pose, Pick & Place is executed using hsrb_interface.