Robot Collision Checking for ROS

A ROS package that wraps around FCL, a popular library in robotics for collision detection and proximity computation. This repository is inspired by MoveIt's approach to interfacing ROS with FCL. However, robot_collision_checking is a lightweight alternative that does not require the entirety of a motion planning framework, like MoveIt, to expose FCL's collision and distance checking capabilities to ROS messages/types.

The robot_collision_checking package can be utilised to perform distance and collision checking of objects by creating and maintaining a collision world and/or by using utility functions (see the API Documentation for more information). This package can handle objects represented as shape_msgs, OctoMaps, and VoxelGrids.

Implementations for the following ROS distros are available on different git branches of this repository: - ROS 1 Noetic on the noetic-devel branch, - ROS 2 Foxy on the foxy branch, and - ROS 2 Humble on the default humble branch.

The package was developed and tested on Ubuntu 20.04 for Noetic/Foxy and Ubuntu 22.04 for Humble. Nevertheless, any operating systems supported by the ROS distros available to this package should also work. We recommend using the default ROS 2 Humble implementation, as this continues to have ongoing support.

In terms of third-party software, this package requires: * FCL * libccd * OctoMap * Eigen

Warning: Older versions of FCL and libccd may already be installed on your system when installing ROS. These versions are incompatible with robot_collision_checking, so please ensure that these libraries are installed according to the Installation section of this repository.

Installation

The following instructions will enable you to build the robot_collision_checking package within a ROS 2 workspace using colcon build (or catkin build if using ROS 1).

FCL and libccd may already be installed on your machine via your ROS distro, but these versions are likely outdated for the current repository's use. You will need to build these libraries from source, as described below.

libccd

In a directory of your choice, run the following commands to build libccd from source: git clone https://github.com/danfis/libccd.git cd libccd && mkdir build && cd build cmake -G "Unix Makefiles" -DENABLE_DOUBLE_PRECISION=ON .. make sudo make install

FCL

Important: Before installing FCL, make sure to have liboctomap-dev installed, e.g., sudo apt install liboctomap-dev as FCL will ignore building OcTree collision geometries otherwise (see Issue #4 for more information).

Once Octomap is installed, run the following commands in a directory of your choice to build an up-to-date version of FCL (i.e., later than version 0.7.0) from source:

git clone https://github.com/flexible-collision-library/fcl.git cd fcl && mkdir build && cd build cmake .. make sudo make install

If there are errors when building robot_collision_checking, such as constants not being found, then you are probably still using an older version of FCL (prior to version 0.7.0).

Workspace Setup

You can now clone the robot_collision_checking package into your ROS workspace's src directory and set it to the appropriate ROS distro implementation, e.g., by cloning as follows: git clone --branch <branch-name> https://github.com/philip-long/robot_collision_checking.git Where <branch-name> is either humble, foxy, or noetic-devel.

Don't forget to install any other system dependencies through rosdep after installing the above libraries, e.g., in the root directory of your ROS workspace run: rosdep install --from-paths src --ignore-src -y

Alternative - Docker Image

If you instead wish to explore the package in a Docker image, there is a Dockerfile available. After installing docker, simply clone the repository or download the Dockerfile and then run: docker build --tag 'robot_collision_checking' . && docker run -it 'robot_collision_checking' bash

The Docker image is preconfigured with all the core libraries for robot_collision_checking (e.g., FCL, libccd, ROS, etc.). After building the image and starting the container, a ROS workspace ros2_ws will have been created and built with all the necessary dependencies. The final step is to source ros2_ws before testing out the package: source /ros2_ws/install/setup.bash

Testing

You can run tests for the robot_collision_checking package as described in this ROS 2 tutorial. First compile the tests: colcon test --ctest-args tests

And then examine the results: colcon test-result --all --verbose

There are seven tests implemented in interface_test.cpp: 1. TransformToFCL: To validate the Eigen::Affine3d to fcl::Transform3d transformation method. 2. AddRemove: Asserts that a variety of shape_msgs and OctoMaps can be added/removed to a collision world. 3. AddRemoveVoxelGrid: Same as "AddRemove", except focused on VoxelGrids. 4. NullPtrCheck: Tests whether the library handles nullptr function arguments correctly. 5. CollisionCheck: Validates that FCL collision-checking capabilities operate correctly for ROS types exposed through this package (e.g., shape_msgs). 6. DistanceCheck: Validates that FCL distance computations are correct for ROS types exposed through this package (e.g., shape_msgs). 7. OctomapCollDistCheck: A combination of tests for FCL collision and distance checking when using robot_collision_checking as an interface to handle OctoMaps as collision geometries.

If all seven tests pass, you can expect minimal output with a summary of "0 failures". If there are any errors, test-result will provide a detailed breakdown of the test(s) that failed and reason why in the terminal output. A test results file will also be stored in the build/robot_collision_checking directory of your ROS workspace (unless configured otherwise -- see ROS tutorial docs).

Examples

A toy example is provided in the examples directory and can be run as follows: ros2 run robot_collision_checking fcl_interface_example In a separate terminal, run an instance of RViz and set the global fixed frame to "world" to visualize the collision world. You can install rviz2 on Debian systems by running: sudo apt install ros-$ROS_DISTRO-rviz2 If everything is set up correctly, you should see a view similar to:

Within the fclinterfaceexample node, a few key pieces of functionality are provided: - First, a collision world composed of different geometric shapes and types (meshes, planes, voxel grids, etc.) is constructed and maintained using the package's interface.

    // Initialize the FCL collision world
    robot_collision_checking::FCLInterfaceCollisionWorld collision_world("world");
    bool success = initCollisionWorld(collision_world);

    bool initCollisionWorld(robot_collision_checking::FCLInterfaceCollisionWorld& world)
    {
        ...
        // Collection of objects to be added to the world
        std::vector<robot_collision_checking::FCLObjectPtr> fcl_objects;
        std::vector<int> object_ids = {0, 1, 2, 3, 4};
        ...
        // Adds the collection of FCL objects to the collision world
        // And returns whether this operation succeeded or failed
        return world.addCollisionObjects(fcl_objects, object_ids);
    }

            where in the initCollisionWorld method, a collection of five FCL objects are added to the world: - Second, the main publishing loop indicates how these different geometric types can be translated into visualization_msgs/Marker messages for visualization in RViz. - Finally, the example shows how the created collision world can be used to check for collisions between its constituent objects.

    std::vector<robot_collision_checking::FCLInterfaceCollisionObjectPtr> world_objects = 
        collision_world.getCollisionObjects();
    for (int i = 0; i < num_objects; /*i++*/)
    {
        auto world_obj = world_objects[i];
        std::string obj_type = world_obj->object->getTypeString();
        ...
        bool is_collision = collision_world.checkCollisionObject(
            world_obj->collision_id, collision_object_ids);
        if (is_collision)
        {
            for (int obj_id : collision_object_ids)
            {
                RCLCPP_INFO(node->get_logger(), "%s with ID %d in collision with object with ID %d", 
                            obj_type.c_str(), world_obj->collision_id, obj_id);
            }
        }
        ...
    }

            The output of the example node prints information about any objects currently in collision (as shown in the code snippet above).

Please refer to the package's API documentation for more information about the code.

While this example only contains static objects, the package also works with dynamic objects. A more extensive use-case of this package that includes dynamic scenarios is provided in constrained_manipulability. Here, the robot_collision_checking interface checks for collisions and distances between environmental objects and a robot manipulator (based on the geometric shapes present in its URDF model).