Android and rosjava

Open source

Random shot of desk

coders and employees from Google have joined forces to create rosjava. This is a pure Java implementation of ROS (from Willow Garage). ROS shouldn’t be seen as a solution for all and everything. So, for what can it be useful? Under other things, it provides:

utilities for robots to make it easier to visualize, compress and communicate data
basic data processing software
drivers for devices of which you might like to address their data (a famous example is the Kinect).

Enough to be really useful! However, there is something that is a little bit a negative point. ROS might be overwhelming if you just start with your first robot or Android project. It can be quite a pain to figure everything out. So here we go with a sort tutorial where we will guide you around (undocumented) issues that will come up.

Documentation

First of all, you definitely need to look at the online documentation of the android_core where you will discover that a lot of documentation online is outdated. The android core for example does not make use of the “rosmake” utility, but of the “gradlew” build utility. If you have not installed ROS yet, feel free to do so. The directions in this blog post will target Ubuntu Precise Penguin (12.10). You can install using sudo aptitude install ros-fuerte-ros. The first thing I have to warn against is installing stuff yourself using mercurial.

Setup

Feel free to setup the system how you want of course. I tend to use multiple Eclipse workspaces to keep code together that belongs together. So, a ~/myworkspace/nonequilibrium for statistical physics code, a ~/myworkspace/android for Java code for Android, a ~/myworkspace/arduino for Arduino related code, a ~/myworkspace/nodejs for node.js code, and naturally for ROS: mkdir ~/myworkspace/ros && cd ~/myworkspace/ros

Please, follow closely the rosws tutorial to create the proper ROS workspace:

sudo apt-get install python-pip

sudo pip install -U rosinstall

rosws init

source setup.bash

And copy that line indeed to your ~/.bashrc file as recommended

Using rosws we can “merge” code from different locations. This will only add references to the given code or repositories, not the code itself.

rosws merge http://android.rosjava.googlecode.com/hg/.rosinstall

rosws merge http://rosjava.googlecode.com/hg/.rosinstall

rosws merge /opt/ros/electric/.rosinstall

Now to get the code actually and check the results:

rosws update

rosws info

# Expected output: workspace: /home/anne/myworkspace/ros ROS_ROOT: /opt/ros/fuerte/ros Localname & nbsp; S SCM Version-Spec UID (Spec) URI (Spec) [https://...] \\--------- &n; bsp; &n; bsp; - ---- ------------ ----------- \------------------------- ros java_core &nb; sp; hg default 7cfb92045e17 code.google.com/p/rosjava /opt/ros/electric/stacks &nb; sp; &nb; sp; &nb; sp; /opt/ros/electric/ros ; ; ; google &nbs; p; &nbs; p; hg default 736ac32a2ede code.google.com/p/rosjava.google android_core &n; bsp; hg default 9327ca89c140 code.google.com/p/rosjava.android

Pay attention that ROS_ROOT is not empty. Eventually check everything using env | grep ROS. You can also use rospack list-names to return a list of all packages.

Hurdle 1: Forgetting to source script

We will try to build our first code:

roscd rosjava_core

roscd: No such package ‘rosjava_core’

If an error like this happens and rosjava is actually in the rosws info list, then you will need to call the script in the ROS workspace.

source setup.sh

roscd rosjava_core

./gradlew install

Hurdle 2: Refer to Android

And then similar the other code:

roscd google

./gradlew install

roscd android_core

./gradlew debug

Execution failed for task ‘:android_acm_serial:updateProject’. A problem occurred starting command ‘android’.

This fails for me because it cannot find android. Now you will think that you only need to update your environmental $PATH, however, it is not just bash that does not find the android binary, but it is ANT, the Java build system. Hence you will need to add this to your ~/.bashrc file:

export PATH=$PATH:/opt/android-sdk- linux_x86/tools

export CLASSPATH=$CLASSPATH:/opt/android-sdk-linux_x86/tools

sudo killall java

Make sure the JVM is dead. Close all neat Java programs such as Eclipse and JabRef beforehand. And now run ./gradlew clean debug again.

On my system it still complains about the fact that a target needs to be specified, but for now we will build one project that does not have so many dependencies:

./gradlew android_tutorial_pubsub:clean android_tutorial_pubsub:debug

Hurdle 3: Set Android target

There is a problem about the absence of the build.xml file on building the files straight from the repository.

Buildfile: build.xml does not exist!

The solution is simple:

cd android_honeycomb_mr2

android list targets

Pick the one that is recent enough for honeycomb, in my case that is number 3.

android update project --target 3 --path .

You can pick android_tutorial_teleop afterwards for example for which you will need to do the same. I am curious in streaming the camera from my smartphone to my computer for now: android_tutorial_camera. Installing on your smartphone is easy by just connecting your smartphone via USB and after the build ant installd (where the d stands for the debug version).

Hurdle 4: Display image

According to the ROS wiki image_view resides apparently in the pipeline package. We also install rxgraph which is a nice utility to visualize our architecture and last, but not least we will need to be able to get compressed images (transport plugins):

sudo aptitude install ros-fuerte-image- common ros-fuerte-image-pipeline

sudo aptitude install ros-fuerte-rx

sudo aptitude install ros-fuerte-image- transport-plugins

Now on your PC start the ROS core and note down your IP address:

roscore

sudo ifconfig

Start the CameraTutorial app on your smartphone and connect to the IP address you just wrote down, e.g. http://10.10.1.133:11311. You can check beforehand if you can reach your PC from your smartphone using the browser on your phone (however, Opera Mobile couldn’t find my server for some reason, so don’t trust such a check too much). Needless to say, use an IP address and not some local name only known to your computer. Finally, start the viewer on the PC side:

rosrun image_view image_view image:=/camera/image compressed

Done! Proof can be found in the form of the picture (from the camera on my smartphone) at the top of this blog post obtained by right-clicking in the image viewer on my PC.