rlinsalata's profile - activity

+1; For those who may have missed the nuance here -- The advantage of doing it this way is that you are actually creating a ROS Node for your executable, and in this way you can tell the process is executing, and possibly control it.

Don't worry, it's harmless.

Frequent question - see: http://answers.ros.org/question/37087/potential-python-problem-in-ubuntu-1204-lts-64bit-and-fuerte/?answer=37088#post-id-37088

If you use ROS_HOSTNAME it will override ROS_IP. However, if you set it to ROS_HOSTNAME=localhost, it will also attach itself to the local-only loopback interface - in other words, you can set it to localhost to cut off any remote access, but it should be a good test to make sure you can launch the nodes, bring-up the turtlebot and that (parts of) ROS are installed correctly. (oh, but you can still ssh into your turtlebot to run stuff remotely)

I'm not the most knowledgeable about networks, but I have been banging my head against them for the past few weeks myself, so I see two other things:

1. Your IP address appears to be behind a private wifi router (maybe, yes?). If you have control of the router, try giving your turtlebot a static IP. This could help a lot.

2. I assume that you eventually may want to access your turtlebot from ROS on other computers, what types of OS are you using on those other computers (Linux, Windows, Mac)? And more importantly, what are your results from pinging the hostname 'turtlebot' and the other "Checks" listed on the Network page Lorenz listed? (1.1, and 1.2)

You say you get trash through the serial port in minicom? During charging (in default, start-up Passive mode), you should be getting clear, regularly repeating battery status messages - something like: "bat: min 124 sec 24 mV 15864 mA -85 deg-C 43". Do you get that in minicom?

There is also "regular" rosserial, along with a (generic?) rosserial_client. On the other hand, an Arduino is just an AVR microcontroller on a standard breakout board. The rosserial page also lists the general protocol, but you'll need the node running to act as a bridge between ROS and the serial connection.

ROS_OS_OVERRIDE does not (seem to) affect rosdep install. I don't know if this is by design or an oversight. However, I got it to work by manually setting the os override option on the command line when calling rosdep:

rosdep resolve <key-name> --os=OS_NAME:OS_VERSION

This option will work with the other rosdep commands such as rosdep install. I did this (to trick rosdep for my Linaro/Ubuntu system):

rosdep install -ay --os=ubuntu:oneiric

I have been going through some similar issues trying to install on PandaBoard from source. The reason someone will need to install from source in Precise, is if they are on an ARM processor (i.e. BeagleBoard, PandaBoard, etc.) - no debian packages for ARM (yet*).

As far as my understanding goes:

1. ros-fuerte-pcl is a 3rd party dependency (<rosdep> key 'pcl'), which has been patched/wrapped into it's own debian package on Willow's wg-debs github (so that it plays nice with ROS).

2. There are a bunch of funky naming things going on with 'pcl'-related packages/stacks/dependencies.

   • EX:

   $ rosdep keys pcl
   pcl
   $ rosdep resolve pcl
   #apt
   ros-fuerte-pcl
   $ rospack find pcl
   /opt/ros/fuerte/share/pcl
   $ roslocate info pcl
   - svn:
   local-name: pcl
   uri: http://svn.pointclouds.org/ros/trunk/perception_pcl_unstable/pcl
   $ roslocate type pcl
   package
   

   Btw, http://svn.pointclouds.org/ros/trunk/perception_pcl_unstable/pcl, doesn't exist
   • Some packages depend on a ROS package named 'pcl' (<depend package="pcl"/>), and some depend on an external package, with the key 'pcl' (<rosdep name="pcl"/> e.g. ros-fuerte-pcl).
   • And, I'm not completely sure, but I believe that the "Missing resource pcl" will be satisfied when you install ros-fuerte-pcl (from package or source). There is a subfolder 'ros' in the ros-fuerte-pcl package whose contents get installed in /opt/ros/fuerte/share/pcl (which is where 'rospack find pcl' points me in my debian package install of Fuerte on Precise x86 machine).
   • Side note: perception_pcl includes a ROS package 'pcl_ros' which I think used to mostly serve to pull in the external dependency.

3. Also, currently rosdep doesn't work for <rosdep> dependencies in stacks (stack.xml). It just ignores them. Thus, if you have to build stacks from source that list their dependencies at the stack level (e.g. ros-fuerte-pcl, opencv) you have to manually go into the stack.xml and install each (debian) package it lists. You may need to resolve those <rosdep> keys into actual package names using rosdep.

4. You will also notice that you're "missing resource protobuf". This is from packages in simulator_gazebo depending on a ROS package called protobuf. The protobuf ROS package/stack exists for Ubuntu Lucid and earlier, which did not distribute the proper version of the library. In Precise, there should be a protobuf debian package (libprotobuf7), even for ARM, which I'd recommend you just apt-get install manually. Then you can just comment out the shell ROS dependency, <depend package="protobuf"/>, from the manifest.xml's in gazebo_words and gazebo. Your missing protobuf resource should disappear in a rosdep check -a

Sorry if that was a bit of an overload; hopefully it shines some light on what's going on. Also, related FYI: http://answers.ros.org/question/35602/ros-fuerte-pandaboard-ubuntu-1204-no-go/

*There might be armel packages (BeagleBoard) possible in the near future, but armhf (PandaBoard) not as much.

The Dexter NXT Wifi Sensor is just that - a sensor. So essentially, you will need to write your own proxy as ipso said, along with a corresponding program on the NXT that will read the "sensor" messages (i.e. whatever custom messages you send over Wifi), and react to control the motors/send back sensors. There is currently no way to connect directly to the NXT's immediate command API connection, as the USB and Bluetooth do.

However, if you're in java already ... there is the Android NXT-Bluetooth SDK.

Is this still the case? I can't seem to find the launch files for the mentioned tf::MessageFilter tutorial, or the Debugging tf tutorial right before it (start_debug_demo.launch), but maybe I'm looking in the wrong place.

A couple people have been noticing/asking about this issue (http://answers.ros.org/question/34304/exception-attributeerror-when-starting-roscore/?answer=34307#post-id-34307).

Apparently it is a Python-internal bug, which we just need to wait for the fix to be released (both tickets closed):

• ROS ticket: https://code.ros.org/trac/ros/ticket/3950
• Python ticket: http://bugs.python.org/issue14308

(Though, I have to wonder if it is related to any of the hiccups I've been encountering in some of the basic python tutorial/example code, e.g. tf, turtle_tf)