This removes the need to specify the body id/link index when retrieving a user data entry.
Additionally, user data can now optionally be set to visual shapes as well.
The following public pybullet APIs have changed (backwards incompatible)
addUserData and getUserDataId
Makes linkIndex parameter optional (default value is -1)
Adds optional visualShapeIndex parameter (default value is -1)
getUserData and removeUserData
Removes required parameters bodyUniqueId and linkIndex
getNumUserData
Removes required bodyUniqueId parameter
getUserDataInfo
Removes required linkIndex parameter
Changes returned tuple from (userDataId, key) to (userDataId, key, bodyUniqueId, linkIndex, visualShapeIndex)
- Limits the maximum number of threads to 64, since btThreadSupportPosix
and btThreadsupportWin32 don't support more than 64 bits at this moment,
due to the use of UINT64 bitmasks. This could be fixed by using
std::bitset or some other alternative.
- Introduces a threadpool class, b3ThreadPool, which is a simple wrapper
around btThreadSupportInterface and uses this instead of the global task
scheduler for parallel raycasting. This is actually quite a bit faster
than the task scheduler (~10-15% in my tests for parallel raycasts),
since the advanced features (parallelFor) are not necessary for the
parallel raycasts.
- Puts 16*1024 of MAX_RAY_INTERSECTION_MAX_SIZE_STREAMING in
parentheses, since it otherwise causes problems with other operators
of equal precedence and introduces a smaller constant for Apple targets.
- Refactors the parallel raycasts code and adds some more profiling.
use b3RaycastBatchAddRays API to enable MAX_RAY_INTERSECTION_BATCH_SIZE_STREAMING num rays.
Old API (b3RaycastBatchAddRay) sticks to 256 rays, MAX_RAY_INTERSECTION_BATCH_SIZE.
(and issue with TaskScheduler/btTaskScheduler.cpp, add JobQueue::exit, call it first, since it uses the m_threadSupport which was deleted before the destrucor was called.
Use a hashmap to store user timers, to avoid allocating many identical strings.
reduce 'm_cooldownTime' from 1000 microseconds to 100 microseconds (overhead in raycast is too large)
If needed, we can expose this cooldown time.
Replace malloc by btAlignedObjectArray (going through Bullet's memory allocator)
To enable the feature, enable the BULLET2_MULTITHREADING option.
Increases the number of rays that can go in a batch request by storing
them in the shared memory stream instead of the shared memory command.
Adds the API b3RaycastBatchSetNumThreads to specify the number of
threads to use for the raycast batch, also adds the argument numThreads
to the pybullet function rayTestBatch.
Rays are distributed among the threads in a greedy fashion there's a shared
queue of work, once a thread finishes its task, it picks the next
available ray from the task. This works better than pre-distributing the
rays among threads, since there's a large variance in computation time per ray.
Some controversial changes:
- Added a pointer to PhysicsClient to the SharedMemoryCommand struct, this
was necessary to keep the C-API the same for b3RaycastBatchAddRay, while
adding the ray to the shared memory stream instead of the command
struct. I think this may be useful to simplify other APIs as well, that
take both a client handle and a command handle.
- Moved #define SHARED_MEMORY_MAX_STREAM_CHUNK_SIZE to
SharedMemoryPublic. This was necessary for the definition of
MAX_RAY_INTERSECTION_BATCH_SIZE.
Extract faces directly from btConvexHullComputer (in initializePolyhedralFeatures), instead of reconstructing them, thanks to Josh Klint in #1654
PyBullet: use initializePolyhedralFeatures for convex hulls and boxes (to allow SAT)
PyBullet: expose setPhysicsEngineParameter(enableSAT=0 or 1) to enable Separating Axis Test based collision detection for convex vs convex/box and convex versus concave triangles (in a triangle mesh).
Split examples/SharedMemory/b3RobotSimulatorClientAPI_NoGUI.* and move to examples/SharedMemory/b3RobotSimulatorClientAPI_NoGUI.cpp and examples/SharedMemory/b3RobotSimulatorClientAPI_NoDirect.cpp
See https://github.com/erwincoumans/pybullet_robots ANYmal.py for an example.
PyBullet: Expose p.setPhysicsEngineParameter(solverResidualThreshold=1e-2) (b3PhysicsParamSetSolverResidualThreshold), increases solver performance a lot
PyBullet: Expose p.setPhysicsEngineParameter(contactSlop) Set it to zero, to avoid issues with restitution.
PyBullet: Expose isNumpyEnabled, return True is PyBullet was compiled with NUMPY support for 'getCameraImage'.
PyBullet: Expose p.ChangeDynamics(objectUid, linkIndex, contactProcessingThreshold), to avoid issues of speculative/predictive contacts with restitution.
See also http://twvideo01.ubm-us.net/o1/vault/gdc2012/slides/Programming%20Track/Vincent_ROBERT_Track_ADifferentApproach.pdf
use p.JOINT_FEEDBACK_IN_JOINT_FRAME if you want the joint feedback expressed in joint frame (instead of link inertial frame)
use p.JOINT_FEEDBACK_IN_WORLD_SPACE if you want the joint feedback in world space coordinates, instead of local link/joint coordinates.
Example: p.setPhysicsEngineParameter(jointFeedbackMode=p.JOINT_FEEDBACK_IN_WORLD_SPACE+p.JOINT_FEEDBACK_IN_JOINT_FRAME)
Use glBufferSubData instead of glMapBuffer/glUnmapBuffer, much faster soft body vertex sync.
Don't use a separate btSoftBodyWorldInfo, use the existing one in btSoftMultiBodyDynamicsWorld.
(due to local convex-triangle collisions causing opposite contact normals, use the pre-computed edge normal)
PyBullet: expose experimental continuous collision detection for maximal coordinate rigid bodies, to prevent tunneling.
use URDF_GLOBAL_VELOCITIES_MB flag in PyBullet loadURDF.
fix robot_bases.py due to new fields in getJointInfo.
backward compabitibility: BulletMJCFImporter, keep creating btMultiSphereShape for MJCF capsules with fromto, instead of shifted btCapsuleShapeZ, unless if CUF_USE_IMPLICIT_CYLINDER is used.
p.loadURDF("r2d2.urdf", flags=p.URDF_USE_IMPLICIT_CYLINDER)
allow to enable/disable deterministicOverlappingPairs through an API
p.setPhysicsEngineParameter(deterministicOverlappingPairs = False)
Use btCylinderShapeZ for URDF cylinder, instead of converting it to a btConvexHullShape.
Implement PyBullet.getCollisionShapeData
Extend PyBullet.getDynamicsInfo / b3GetDynamicsInfo, remove flag (don't rely on API returning a fixed number of elements in a list!)
Extend PyBullet.getJointInfo: add parentIndex
Option to de/serialize btPersistentContactManifolds and fix lossy conversion during btMultiBody de/serialization of base world transform
(serialize the quaternion, not the converted 3x3 matrix)
There are still several caches not taken into account, and btMultiBody links/constraints are not deserialized yet etc.
See examples\pybullet\examples\saveRestoreState.py for an example.
render the inertia boxes in examples/pybullet/examples/quadruped.py and examples/pybullet/examples/reset_dynamic_info.py
fix an issue where the original margin (0.04) was used to compute the inertia, instead of latest margin
Over time, some of those functions can be moved into separate plugins, similar to vrSyncPlugin (either statically/dynamically linked plugins or dynamically loaded at run-time)
vr_kuka_control.py: control all joints, use analogue button to close gripper
remove some debug warnings/prints
pybullet, avoid crash in changeUserConstraint if not passing a [list]
allow some gym environments (pybullet_pendulum and locomotors) to re-use an existing physics client connection.
option to disable tinyrenderer, use p.configureDebugVisualizer(p.COV_ENABLE_TINY_RENDERER,0)
also make sure to use p.configureDebugVisualizer(p.COV_ENABLE_RENDERING,0) before loadURDF, and enable rendering afterwards using p.configureDebugVisualizer(p.COV_ENABLE_RENDERING,1)
reorder 2 loops, making the flip texels twice as fast (cache coherency),
single memcpy of entire texture in tinyrenderer, instead of per-pixel copy (memory layout is the same)
add lots of B3_PROFILE timings, to see where time is going
pass events (keyboard, mouse, vr controllers etc) to the plugin, and clear them after the tick callback, so that it doesn't interfere with Python 'getEvents'
add option to recompute forward kinematics, to be consistent with link velocities in pybullet.getLinkState (..., computeForwardKinematics=0/1), thanks to Jeff Bingham for bringing up this inconsistency
The server command processor actually didn't do anything with
the local point that was passed along with the calculateJacobian
command. Added in the necessary bit of math to return the
corresponding jacobian.
Also, fixed a typo in pybullet that was returning the same
jacobian for translation and rotation.
* Add the calculateJacobian method to the pybullet API.
* Adjust the shared memory interface to handle fixed/floating bases
in the calculateJacobian method.
* Fix a few comments.
Add preTickPluginCallback/postTickPluginCallback
User pointer for b3PluginContext, to store objects (class/struct instances)
Pass ints and floats as optional argument for plugin executePluginCommand
use debug visualizer camera viewmatrix/projection matrix if not provided, in ER_BULLET_HARDWARE_OPENGL mode.
fix broken changeRGBAColor implementation, thanks to Laura for the report!
Reduce size of command-logfile (used in PhysicsServer (Logging) and PhysicsServer (Log Replay))
Make Bullet3Common and BulletInverseDynamics part of core Bullet libraries (not optional)
fix in indexing for maximal coordinates (unused by default, still experimental, requires many iterations for Minitaur due to extreme mass-ratio, hence use of reduces/generalized coordinates)
modify quadruped.py to test maximal coordinates
wrap angular servo (positional) target within [-PI,PI] in btGeneric6DofSpring2Constraint
add 'j' key to show body frames in wireframe/debug mode
See quadruped.py for an example:
p.startStateLogging(p.STATE_LOGGING_GENERIC_ROBOT, "genericlogdata.bin", maxLogDof = 16, logFlags = p.STATE_LOG_JOINT_TORQUES)
Thanks to JulianYG, in pull request https://github.com/bulletphysics/bullet3/pull/1273
