repository·master·Indexed 21 days ago
https://github.com/rdiankov/openraveA robotics framework featuring tools for motion planning, collision checking, and trajectory management. Includes documentation on optimizing planning speed via joint resolution tuning, planner parameters, and collision checking techniques, as well as integration with the Breathe Sphinx extension for Doxygen XML output. The repository also contains third-party libraries including ANN, FLANN, ConvexDecomposition, cr-libm, ivcon, PCRE, and zlib.
OpenRAVE is a simulation and analysis environment designed for testing, developing, and deploying motion planning algorithms in real-world robotics applications.
Key features include:
ANN (Approximate Nearest Neighbors) is a C++ library designed for both exact and approximate nearest neighbor searching across various dimensions. Beyond searching, it serves as a testbed for:
For detailed usage, refer to the 'ANN Programming Manual' provided with the software distribution.
Convex Decomposition is the process of subdividing an arbitrarily complex triangle mesh into a collection of discrete compound pieces, where each piece is represented as a convex hull. This approximates the original shape of the object.
Use Case: Most physics engines cannot efficiently treat arbitrary triangle meshes as dynamic objects. Using raw meshes for dynamic simulations incurs significant performance and memory penalties. By breaking a complex mesh into multiple convex components, you can significantly improve performance for dynamic simulations.
While all Environment methods are multi-thread safe, methods belonging to KinBody, Robot, Controller, Planner, etc., are not thread-safe. To perform operations on these objects safely, you must lock the environment.
Locking is performed using Environment.Lock(dolock). It is recommended to use the Python with statement for scoped locking:
env = Environment()
# initialization code
with env:
# environment is now locked
env.CheckCollision(...)OpenRAVE is designed as a plugin-based simulation and planning framework. Instead of a monolithic codebase, it uses a distributed architecture where functionality—such as planning algorithms, robot control, and sensing subsystems—is implemented as plugins that can be dynamically loaded at run-time.
This architecture allows you to:
OpenRAVE extends the COLLADA 1.5 specification to include robotics-specific data like manipulators, sensors, and collision data. To provide this information, use the <extra> tag with the OpenRAVE technique profile.
There is a one-to-one correspondence between OpenRAVE interface types and COLLADA tags:
<articulated_system><sensor><manipulator>To enable exact trajectory collision checking, you must implement the DistanceCheckerBase interface. This implementation must return a lower bound on the $L_{\infty}$ distance between a configuration and the obstacle boundary in C-space.
Once implemented, pass both your FeasibilityCheckerBase and your DistanceCheckerBase to the overloaded RampFeasibilityChecker constructor. When using this variant, you do not need to implement FeasibilityCheckerBase::SegmentFeasible.
OpenRAVE supports the COLLADA format for specifying robots and scenes, including robot-specific extensions.
For geometric consistency, ensure you follow the project's geometric_conventions regarding coordinate systems and scale.
The standard <formula>/<technique_common> in COLLADA only supports a single equation for a joint value. To support complex kinematics with multiple degrees of freedom per joint, use the OpenRAVE technique. This allows you to specify partial derivatives of the position equation, which are used to compute velocities, accelerations, and Jacobians.
Within a <technique profile="OpenRAVE"> block, use multiple <equation> elements:
type="position": Defines the position equation in MathML.type="first_partial": Defines the first partial derivative. Requires the target attribute to specify the variable being differentiated.type="second_partial": Defines the second partial derivative. Requires the target attribute.<technique profile="OpenRAVE">
<equation type="position">
<math>
<apply>
<plus/>
<apply>
<times/>
<cn>0.333330</cn>
<csymbol encoding="COLLADA">kmodel1/joint0</csymbol>
</apply>
<cn>0.872700</cn>
</apply>
</math>
</equation>
<equation type="first_partial" target="kmodel1/joint0">
<math>
<cn>0.333330</cn>
</math>
</equation>
</technique>All HTML templates for the website are located in the openrave_website/templates directory. The site uses Django's template language.
When writing templates, follow these rules for internationalization:
{% trans %} or {% blocktrans %} tags.