peci1's profile - activity

I have a multi-link robot SDF in Gazebo 2.1. It is a simple "static" model with just inertia and collision elements defined, no transmission elements or motors specified.

When I let the robot fall down on a plane, everything is ok.

However, I need to use a terrain model obtained from point cloud triangulation. Such terrain contains lots of small triangles with varying normals and errors (because of precision of the measurements and triangulation). When I let the robot fall on this terrain, it instantly begins to fall apart (even if the terrain is quite planar/the normals do not differ much).

My workaround so far is to preprocess the SDF to remove all joints (in different words - the links remain, just the joints are static), so that for every joint config I have a different SDF model. Highly ineffective, I know. Luckily, by this modification, the model stops falling apart and behaves in an expected way.

Do you have an idea on how to have a model with non-static joints that doesn't fall apart?

Usually, this happens if the Gazebo model has fixed joints and there are some parts with highly different inertia scales

Is there a place where one could quickly decide which physics engine to use knowing his requirements (e.g. dynamic changes to models during runtime, joint friction, whatever).

It happens that some functionality is only implemented for ODE, but not the other engines. It would be nice to see all these "TODOs" in one place.

For (Ignition) Gazebo, the list is here: https://gazebosim.org/docs/garden/comparison .

When 2 bodies are colliding and I set Gazebo (2.2) to show contacts, there is a blue ball and a line for each contact point. The lines are either red or green.

What do the colors mean?

My estimate is that green is for normal collisions, whereas red is for penetration (which should not occur with rigid bodies).

I assume these two cases can be distinguished using the contact depth information provided in the contact message, and that there is a threshold to tell if the line should be green or red. Where is the definition of the threshold? Or am I completely wrong?

Xvfb is much easier to use. xvfb-run -a gazebo and that's it. It also doesn't need root privileges and doesn't need a re

Xvfb is much easier to use. xvfb-run -a gazebo and that's it.

You need to make sure the containers can see each other (be in the same IP subnet) and that the hostnames they report ca

You need to make sure the containers can see each other (be in the same IP subnet) and that the hostnames they report ca

Do you intend to keep the loaded .world file the same and only change the robots spawned in the sim? AFAIK changing the

Hi, there's not exactly a command, but the API is there. You can find it in the TrackController system. Some more relate

Synchronous publishing in Ignition Gazebo in PreUpdate? I'm writing an Ignition Gazebo model plugin and I'd like to comm

I also faced this issue recently. In my case, I was trying to load an SDF generated for Ignition Gazebo in Gazebo Classi

It works if the collision has no explicit name in URDF or if the name of the collision is $LINKNAME_collision. Any other

If anybody has a copy of that answer, please post it here. @jose-luis-rivero ?

The wiki with instructions is a bit outdated. It says it is for building gazebo 6. I now successfully build gazebo8 br

It's an inconsistency in the Protobuf archive provided by OSRF. This path from instructions protobuf-2.6.0-win64-vc12

See http://answers.gazebosim.org/question/16234/how-to-install-ignition-math3-on-windows-10/ . You're building default b

List of per-physics engine capabilities/compatibility/known incompatibilities Is there a place where one could quickly d

I know I'm resurrecting an old thread, but maybe others might still be interested.

Setting fdir1would be a solution if it worked correctly. For some reason, the direction is not treated as relative to the link, but in world coordinates, so you'd need to update it in every frame to the correct world direction. See https://bitbucket.org/osrf/gazebo/iss... .

Also, an omni-wheel needs the most precise friction computations that are available to work properly. Recently, the friction_cone model was added, which helped a lot in my case:

physics::PhysicsEnginePtr physics = world->GetPhysicsEngine();
const std::string frictionModel = "cone_model";
physics->SetParam("friction_model", frictionModel);

Last, a bit unrelated, but nevertheless, now there's (my) PR to add proper support of tracked vehicles. Everybody's welcome to try it out. https://bitbucket.org/osrf/gazebo/pul...

I agree the "ROS way" would be the easiest one.

I tried using the C++ API of tensorflow in Gazebo, and there you are stopped by the fact that Tensorflow requires libprotobuf >= 3, but Gazebo requires libprotobuf < 3. This is generally a big problem if you want to link your app. I'm right now trying some tricks with PIMPL and static linking, so let's see if it's at least possible with a lot of effort.

When I read through the tutorials Using A URDF In Gazebo and Using roslaunch files to Spawn Models I have the impression that it is possible to run Gazebo with ROS without using Gazebo's model database.

To explain, I need to work with some models from ROS using Gazebo. In my view, adding the models to a local database doesn't make sense, because the users of my code just want to git clone and catkin_make and that's all.

So I searched for a way to specify paths to mesh files from my URDF relatively to the package they are contained in. Using the file:// protocol doesn't seem to be a good option, since its resolution is dependent on the working directory.

Somewhere I've found that a package:// protocol can be used, and it actually works in my models. However, there doesn't seem to be any documentation for that. Do you know of docs page that would explain the details and prerequisites for using that?

I've thought about another way to use my models without the need to install them locally. I could just add the package path to GAZEBO_RESOURCE_PATH. But where to do this? Should I use <env> tag in the launch file? Do you have any experience with that?