Cellphone robot ROS Package quick tutorial

Welcome to cellphone robot project! This is part of Project Cellphone robot and is mainly focused on the some extended functionalities, hardware control, as well as message-passing between the android application and the ROS system resided on any linux system. In our project we decide to use Raspberry Pi 3 Model B, a single-board computer with considerate computation power. Read wiki for concrete details.

Structure of Package

Interaction with Android

Step 1: Install ROS

• Follow the link and configure the ROS environment. For instance you can build the catkin works space in ~/catkin_ws.
• Type the following line in your terminal.

$ cd ~/catkin_ws/src
$ git clone https://github.com/wang3303/ros_cellphonerobot.git
$ cd ~/catkin_ws
$ catkin_make

• Note: ROS_PACKAGE_PATH environment variable includes the directory you're in. You should see something similar as below in your terminal.

$ echo $ROS_PACKAGE_PATH
/home/youruser/catkin_ws/src:/opt/ros/kinetic/share

Step 2: Run the robot system using the following line:

$ roslaunch ros_cellphonerobot robot.launch

This will bring all the nodes online.

Step 3: Change profiles in rosparam and set up your robot

You can modify the profile for your robot in /ros_cellphonerobot/rosparam/profile.yaml. The explanation for this file could be found here.

Step 4: Action publishing

You can run the following line to publish keys to /action topic for debugging. For instance, w could stand for moving forward. a could stand for moving backward.

$ rosrun ros_cellphonerobot key_publisher.py

Step 5 (Optional): Image classification

• Install Tensorflow
• Uncomment the node image_classify.py in ros_cellphonerobot/. You can publish images to topic /inception

• Important Note: ros_inception.py will download Neural Network weights and models the first time it runs. Hence, wait patiently until ros_inception.py successfully download necessary files. You can monitor downloading progress by output ros_inception.py to screen after modifying the launch file.

<node pkg="ros_cellphonerobot" name="image_classify" type="image_classify.py" output="screen"/>

• Make sure you pass argument --model_dir and --num_top_predictions to this node as instructed here.