AuroraMiddleware/bullet3
1
0
Fork 0
mirror of https://github.com/bulletphysics/bullet3 synced 2024-12-14 13:50:04 +00:00
Code Issues Releases Wiki Activity
d49e3d787a
bullet3/examples/SharedMemory/PhysicsClientSharedMemory.cpp

966 lines
34 KiB
C++
Raw Normal View History

make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
#include "PhysicsClientSharedMemory.h"
#include "PosixSharedMemory.h"
#include "Win32SharedMemory.h"
#include "LinearMath/btAlignedObjectArray.h"
#include "LinearMath/btVector3.h"
sync repo
2015-11-11 20:44:26 +00:00
#include <string.h>
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
#include "Bullet3Common/b3Logging.h"
#include "../Utils/b3ResourcePath.h"
#include "../../Extras/Serialize/BulletFileLoader/btBulletFile.h"
#include "../../Extras/Serialize/BulletFileLoader/autogenerated/bullet.h"
#include "SharedMemoryBlock.h"
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
#include "BodyJointInfoUtility.h"
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
struct BodyJointInfoCache
{
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
std::string m_baseName;
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
btAlignedObjectArray<b3JointInfo> m_jointInfo;
};
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
struct PhysicsClientSharedMemoryInternalData {
SharedMemoryInterface* m_sharedMemory;
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
bool m_ownsSharedMemory;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
SharedMemoryBlock* m_testBlock1;
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
btHashMap<btHashInt,BodyJointInfoCache*> m_bodyJointMap;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
btAlignedObjectArray<TmpFloat3> m_debugLinesFrom;
btAlignedObjectArray<TmpFloat3> m_debugLinesTo;
btAlignedObjectArray<TmpFloat3> m_debugLinesColor;
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
int m_cachedCameraPixelsWidth;
int m_cachedCameraPixelsHeight;
btAlignedObjectArray<unsigned char> m_cachedCameraPixelsRGBA;
btAlignedObjectArray<float> m_cachedCameraDepthBuffer;
add segmentation mask rendering to TinyRenderer and shared memory API similar to the zbuffer, but storing the object index (int) instead of float depth
2016-08-11 21:55:30 +00:00
btAlignedObjectArray<int> m_cachedSegmentationMaskBuffer;
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
btAlignedObjectArray<b3ContactPointData> m_cachedContactPoints;
pybullet getOverlappingObjects added: report all object unique ids within a world space bounding box (aabb)
2016-11-10 19:22:22 +00:00
btAlignedObjectArray<b3OverlappingObject> m_cachedOverlappingObjects;
first version of 'getVisualShapeData' to get visual shape information to allow external renderer with pybullet and shared memory robotics API b3InitRequestVisualShapeInformation/b3GetVisualShapeInformation in shared memory API
2016-10-19 05:05:28 +00:00
btAlignedObjectArray<b3VisualShapeData> m_cachedVisualShapes;
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
btAlignedObjectArray<int> m_bodyIdsRequestInfo;
SharedMemoryStatus m_tempBackupServerStatus;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
SharedMemoryStatus m_lastServerStatus;
int m_counter;
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
bool m_isConnected;
bool m_waitingForServer;
bool m_hasLastServerStatus;
int m_sharedMemoryKey;
bool m_verboseOutput;
PhysicsClientSharedMemoryInternalData()
: m_sharedMemory(0),
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
m_ownsSharedMemory(false),
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
m_testBlock1(0),
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
m_counter(0),
m_cachedCameraPixelsWidth(0),
m_cachedCameraPixelsHeight(0),
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
m_isConnected(false),
m_waitingForServer(false),
m_hasLastServerStatus(false),
m_sharedMemoryKey(SHARED_MEMORY_KEY),
m_verboseOutput(false) {}
void processServerStatus();
bool canSubmitCommand() const;
};
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
int PhysicsClientSharedMemory::getNumBodies() const
{
return m_data->m_bodyJointMap.size();
}
sync repo
2015-11-11 20:44:26 +00:00
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
int PhysicsClientSharedMemory::getBodyUniqueId(int serialIndex) const
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
{
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
if ((serialIndex >= 0) && (serialIndex < getNumBodies()))
{
return m_data->m_bodyJointMap.getKeyAtIndex(serialIndex).getUid1();
}
return -1;
}
bool PhysicsClientSharedMemory::getBodyInfo(int bodyUniqueId, struct b3BodyInfo& info) const
{
BodyJointInfoCache** bodyJointsPtr = m_data->m_bodyJointMap[bodyUniqueId];
if (bodyJointsPtr && *bodyJointsPtr)
{
BodyJointInfoCache* bodyJoints = *bodyJointsPtr;
info.m_baseName = bodyJoints->m_baseName.c_str();
return true;
}
return false;
}
int PhysicsClientSharedMemory::getNumJoints(int bodyUniqueId) const
{
BodyJointInfoCache** bodyJointsPtr = m_data->m_bodyJointMap[bodyUniqueId];
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
if (bodyJointsPtr && *bodyJointsPtr)
{
BodyJointInfoCache* bodyJoints = *bodyJointsPtr;
return bodyJoints->m_jointInfo.size();
}
return 0;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
bool PhysicsClientSharedMemory::getJointInfo(int bodyUniqueId, int jointIndex, b3JointInfo& info) const
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
{
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
BodyJointInfoCache** bodyJointsPtr = m_data->m_bodyJointMap[bodyUniqueId];
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
if (bodyJointsPtr && *bodyJointsPtr)
{
BodyJointInfoCache* bodyJoints = *bodyJointsPtr;
work-in-progress add UDP network connection for physics client <-> server. also set spinning friction in rolling friction demo (otherwise objects may keep on spinning forever)
2016-11-04 20:15:10 +00:00
if ((jointIndex >= 0) && (jointIndex < bodyJoints->m_jointInfo.size()))
{
info = bodyJoints->m_jointInfo[jointIndex];
return true;
}
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
}
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
return false;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
}
PhysicsClientSharedMemory::PhysicsClientSharedMemory()
{
m_data = new PhysicsClientSharedMemoryInternalData;
#ifdef _WIN32
m_data->m_sharedMemory = new Win32SharedMemoryClient();
#else
m_data->m_sharedMemory = new PosixSharedMemory();
#endif
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
m_data->m_ownsSharedMemory = true;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
}
PhysicsClientSharedMemory::~PhysicsClientSharedMemory() {
if (m_data->m_isConnected) {
disconnectSharedMemory();
}
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
if (m_data->m_ownsSharedMemory)
{
delete m_data->m_sharedMemory;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
delete m_data;
}
void PhysicsClientSharedMemory::setSharedMemoryKey(int key) { m_data->m_sharedMemoryKey = key; }
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
void PhysicsClientSharedMemory::setSharedMemoryInterface(class SharedMemoryInterface* sharedMem)
{
if (m_data->m_sharedMemory && m_data->m_ownsSharedMemory)
{
delete m_data->m_sharedMemory;
}
m_data->m_ownsSharedMemory = false;
m_data->m_sharedMemory = sharedMem;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
void PhysicsClientSharedMemory::disconnectSharedMemory() {
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
if (m_data->m_isConnected && m_data->m_sharedMemory) {
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
m_data->m_sharedMemory->releaseSharedMemory(m_data->m_sharedMemoryKey, SHARED_MEMORY_SIZE);
}
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
m_data->m_isConnected = false;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
}
bool PhysicsClientSharedMemory::isConnected() const { return m_data->m_isConnected; }
bool PhysicsClientSharedMemory::connect() {
/// server always has to create and initialize shared memory
bool allowCreation = false;
m_data->m_testBlock1 = (SharedMemoryBlock*)m_data->m_sharedMemory->allocateSharedMemory(
m_data->m_sharedMemoryKey, SHARED_MEMORY_SIZE, allowCreation);
if (m_data->m_testBlock1) {
if (m_data->m_testBlock1->m_magicId != SHARED_MEMORY_MAGIC_NUMBER) {
b3Error("Error: please start server before client\n");
m_data->m_sharedMemory->releaseSharedMemory(m_data->m_sharedMemoryKey,
SHARED_MEMORY_SIZE);
m_data->m_testBlock1 = 0;
return false;
} else {
if (m_data->m_verboseOutput) {
b3Printf("Connected to existing shared memory, status OK.\n");
}
m_data->m_isConnected = true;
}
} else {
b3Error("Cannot connect to shared memory");
return false;
}
return true;
}
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
///todo(erwincoumans) refactor this: merge with PhysicsDirect::processBodyJointInfo
void PhysicsClientSharedMemory::processBodyJointInfo(int bodyUniqueId, const SharedMemoryStatus& serverCmd)
{
bParse::btBulletFile bf(
&this->m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[0],
work-in-progress add UDP network connection for physics client <-> server. also set spinning friction in rolling friction demo (otherwise objects may keep on spinning forever)
2016-11-04 20:15:10 +00:00
serverCmd.m_numDataStreamBytes);
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
bf.setFileDNAisMemoryDNA();
bf.parse(false);
BodyJointInfoCache* bodyJoints = new BodyJointInfoCache;
m_data->m_bodyJointMap.insert(bodyUniqueId,bodyJoints);
for (int i = 0; i < bf.m_multiBodies.size(); i++)
{
int flag = bf.getFlags();
if ((flag & bParse::FD_DOUBLE_PRECISION) != 0)
{
Bullet::btMultiBodyDoubleData* mb =
(Bullet::btMultiBodyDoubleData*)bf.m_multiBodies[i];
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
bodyJoints->m_baseName = mb->m_baseName;
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
addJointInfoFromMultiBodyData(mb,bodyJoints, m_data->m_verboseOutput);
} else
{
Bullet::btMultiBodyFloatData* mb =
(Bullet::btMultiBodyFloatData*)bf.m_multiBodies[i];
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
bodyJoints->m_baseName = mb->m_baseName;
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
addJointInfoFromMultiBodyData(mb,bodyJoints, m_data->m_verboseOutput);
}
}
if (bf.ok()) {
if (m_data->m_verboseOutput)
{
b3Printf("Received robot description ok!\n");
}
} else
{
b3Warning("Robot description not received");
}
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
const SharedMemoryStatus* PhysicsClientSharedMemory::processServerStatus() {
SharedMemoryStatus* stat = 0;
if (!m_data->m_testBlock1) {
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
m_data->m_lastServerStatus.m_type = CMD_SHARED_MEMORY_NOT_INITIALIZED;
return &m_data->m_lastServerStatus;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
}
if (!m_data->m_waitingForServer) {
return 0;
}
Allow InProcessExampleBrowser to use a malloc allocated memory block, instead of system shared memory. Make shared memory client/server a bit more robust, in case the server is terminated early.
2016-03-10 22:36:46 +00:00
if (m_data->m_testBlock1->m_magicId != SHARED_MEMORY_MAGIC_NUMBER)
{
m_data->m_lastServerStatus.m_type = CMD_SHARED_MEMORY_NOT_INITIALIZED;
return &m_data->m_lastServerStatus;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
if (m_data->m_testBlock1->m_numServerCommands >
m_data->m_testBlock1->m_numProcessedServerCommands) {
btAssert(m_data->m_testBlock1->m_numServerCommands ==
m_data->m_testBlock1->m_numProcessedServerCommands + 1);
const SharedMemoryStatus& serverCmd = m_data->m_testBlock1->m_serverCommands[0];
m_data->m_lastServerStatus = serverCmd;
EnumSharedMemoryServerStatus s = (EnumSharedMemoryServerStatus)serverCmd.m_type;
// consume the command
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
switch (serverCmd.m_type) {
case CMD_CLIENT_COMMAND_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Server completed command");
}
break;
}
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
case CMD_SDF_LOADING_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Server loading the SDF OK\n");
}
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
break;
}
work-in-progress add UDP network connection for physics client <-> server. also set spinning friction in rolling friction demo (otherwise objects may keep on spinning forever)
2016-11-04 20:15:10 +00:00
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
case CMD_URDF_LOADING_COMPLETED: {
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
if (m_data->m_verboseOutput) {
b3Printf("Server loading the URDF OK\n");
}
work-in-progress add UDP network connection for physics client <-> server. also set spinning friction in rolling friction demo (otherwise objects may keep on spinning forever)
2016-11-04 20:15:10 +00:00
if (serverCmd.m_numDataStreamBytes > 0) {
sync repo
2015-11-11 20:44:26 +00:00
bParse::btBulletFile bf(
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
this->m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor,
work-in-progress add UDP network connection for physics client <-> server. also set spinning friction in rolling friction demo (otherwise objects may keep on spinning forever)
2016-11-04 20:15:10 +00:00
serverCmd.m_numDataStreamBytes);
sync repo
2015-11-11 20:44:26 +00:00
bf.setFileDNAisMemoryDNA();
bf.parse(false);
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
int bodyUniqueId = serverCmd.m_dataStreamArguments.m_bodyUniqueId;
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
BodyJointInfoCache* bodyJoints = new BodyJointInfoCache;
expose getNumBodies, getBodyUniqueId, getBodyInfo (char* baseName) to shared memory API and pybullet., making it easier to serialize the state of the world.
2016-09-27 19:13:45 +00:00
m_data->m_bodyJointMap.insert(bodyUniqueId,bodyJoints);
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
sync repo
2015-11-11 20:44:26 +00:00
for (int i = 0; i < bf.m_multiBodies.size(); i++) {
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
int flag = bf.getFlags();
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
if ((flag & bParse::FD_DOUBLE_PRECISION) != 0) {
Bullet::btMultiBodyDoubleData* mb =
sync repo
2015-11-11 20:44:26 +00:00
(Bullet::btMultiBodyDoubleData*)bf.m_multiBodies[i];
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
addJointInfoFromMultiBodyData(mb,bodyJoints, m_data->m_verboseOutput);
} else
{
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
Bullet::btMultiBodyFloatData* mb =
sync repo
2015-11-11 20:44:26 +00:00
(Bullet::btMultiBodyFloatData*)bf.m_multiBodies[i];
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
addJointInfoFromMultiBodyData(mb,bodyJoints, m_data->m_verboseOutput);
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
}
}
sync repo
2015-11-11 20:44:26 +00:00
if (bf.ok()) {
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
if (m_data->m_verboseOutput) {
b3Printf("Received robot description ok!\n");
}
} else {
b3Warning("Robot description not received");
}
}
break;
}
case CMD_DESIRED_STATE_RECEIVED_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Server received desired state");
}
break;
}
case CMD_STEP_FORWARD_SIMULATION_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Server completed step simulation");
}
break;
}
case CMD_URDF_LOADING_FAILED: {
if (m_data->m_verboseOutput) {
b3Printf("Server failed loading the URDF...\n");
}
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
break;
}
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
case CMD_BODY_INFO_COMPLETED:
{
if (m_data->m_verboseOutput) {
b3Printf("Received body info\n");
}
int bodyUniqueId = serverCmd.m_dataStreamArguments.m_bodyUniqueId;
processBodyJointInfo(bodyUniqueId, serverCmd);
break;
}
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
case CMD_SDF_LOADING_FAILED: {
if (m_data->m_verboseOutput) {
b3Printf("Server failed loading the SDF...\n");
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
break;
}
case CMD_BULLET_DATA_STREAM_RECEIVED_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Server received bullet data stream OK\n");
}
break;
}
case CMD_BULLET_DATA_STREAM_RECEIVED_FAILED: {
if (m_data->m_verboseOutput) {
b3Printf("Server failed receiving bullet data stream\n");
}
break;
}
case CMD_ACTUAL_STATE_UPDATE_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Received actual state\n");
}
SharedMemoryStatus& command = m_data->m_testBlock1->m_serverCommands[0];
int numQ = command.m_sendActualStateArgs.m_numDegreeOfFreedomQ;
int numU = command.m_sendActualStateArgs.m_numDegreeOfFreedomU;
if (m_data->m_verboseOutput) {
b3Printf("size Q = %d, size U = %d\n", numQ, numU);
}
char msg[1024];
{
sprintf(msg, "Q=[");
for (int i = 0; i < numQ; i++) {
if (i < numQ - 1) {
sprintf(msg, "%s%f,", msg,
command.m_sendActualStateArgs.m_actualStateQ[i]);
} else {
sprintf(msg, "%s%f", msg,
command.m_sendActualStateArgs.m_actualStateQ[i]);
}
}
sprintf(msg, "%s]", msg);
}
if (m_data->m_verboseOutput) {
b3Printf(msg);
}
{
sprintf(msg, "U=[");
for (int i = 0; i < numU; i++) {
if (i < numU - 1) {
sprintf(msg, "%s%f,", msg,
command.m_sendActualStateArgs.m_actualStateQdot[i]);
} else {
sprintf(msg, "%s%f", msg,
command.m_sendActualStateArgs.m_actualStateQdot[i]);
}
}
sprintf(msg, "%s]", msg);
}
if (m_data->m_verboseOutput) {
b3Printf(msg);
}
if (m_data->m_verboseOutput) {
b3Printf("\n");
}
break;
}
case CMD_RESET_SIMULATION_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("CMD_RESET_SIMULATION_COMPLETED clean data\n");
}
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
m_data->m_debugLinesFrom.clear();
m_data->m_debugLinesTo.clear();
m_data->m_debugLinesColor.clear();
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
for (int i=0;i<m_data->m_bodyJointMap.size();i++)
{
BodyJointInfoCache** bodyJointsPtr = m_data->m_bodyJointMap.getAtIndex(i);
if (bodyJointsPtr && *bodyJointsPtr)
{
sync repo
2015-11-11 20:44:26 +00:00
BodyJointInfoCache* bodyJoints = *bodyJointsPtr;
for (int j=0;j<bodyJoints->m_jointInfo.size();j++) {
if (bodyJoints->m_jointInfo[j].m_jointName)
{
free(bodyJoints->m_jointInfo[j].m_jointName);
}
if (bodyJoints->m_jointInfo[j].m_linkName)
{
free(bodyJoints->m_jointInfo[j].m_linkName);
}
}
initial support for multiple robots in shared memory API
2015-10-13 18:32:25 +00:00
delete (*bodyJointsPtr);
}
}
m_data->m_bodyJointMap.clear();
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
break;
}
case CMD_DEBUG_LINES_COMPLETED: {
if (m_data->m_verboseOutput) {
b3Printf("Success receiving %d debug lines",
serverCmd.m_sendDebugLinesArgs.m_numDebugLines);
}
int numLines = serverCmd.m_sendDebugLinesArgs.m_numDebugLines;
float* linesFrom =
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
(float*)&m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[0];
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
float* linesTo =
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
(float*)(&m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[0] +
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
numLines * 3 * sizeof(float));
float* linesColor =
add physics server loopback (both client and server in the same process, using shared memory) add physics server direct (client and server in the same process, directly processing commands without shared memory transport mechanism)
2015-11-23 04:50:32 +00:00
(float*)(&m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[0] +
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
2 * numLines * 3 * sizeof(float));
m_data->m_debugLinesFrom.resize(serverCmd.m_sendDebugLinesArgs.m_startingLineIndex +
numLines);
m_data->m_debugLinesTo.resize(serverCmd.m_sendDebugLinesArgs.m_startingLineIndex +
numLines);
m_data->m_debugLinesColor.resize(
serverCmd.m_sendDebugLinesArgs.m_startingLineIndex + numLines);
for (int i = 0; i < numLines; i++) {
TmpFloat3 from = CreateTmpFloat3(linesFrom[i * 3], linesFrom[i * 3 + 1],
linesFrom[i * 3 + 2]);
TmpFloat3 to =
CreateTmpFloat3(linesTo[i * 3], linesTo[i * 3 + 1], linesTo[i * 3 + 2]);
TmpFloat3 color = CreateTmpFloat3(linesColor[i * 3], linesColor[i * 3 + 1],
linesColor[i * 3 + 2]);
m_data
->m_debugLinesFrom[serverCmd.m_sendDebugLinesArgs.m_startingLineIndex + i] =
from;
m_data->m_debugLinesTo[serverCmd.m_sendDebugLinesArgs.m_startingLineIndex + i] =
to;
m_data->m_debugLinesColor[serverCmd.m_sendDebugLinesArgs.m_startingLineIndex +
i] = color;
}
break;
}
Allow to request the state of a rigid body (position, orientation, lin/ang velocity) through shared memory API PhysicsClientC_API: b3RequestActualStateCommandInit requires a body unique Id as second argument
2015-10-27 22:46:13 +00:00
case CMD_RIGID_BODY_CREATION_COMPLETED:
{
break;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
case CMD_DEBUG_LINES_OVERFLOW_FAILED: {
b3Warning("Error receiving debug lines");
m_data->m_debugLinesFrom.resize(0);
m_data->m_debugLinesTo.resize(0);
m_data->m_debugLinesColor.resize(0);
break;
}
Expose a better API to allow any render engine to be used for the physics simulation when loading URDF/SDF files. See bullet3/examples/Importers/ImportURDFDemo/DefaultVisualShapeConverter.h Give the kuka_iiwa/model.urdf some blue color, not just orange, to mimick the original a bit better Preparation for the CMD_CAMERA_IMAGE_COMPLETED command, to expose a virtual camera to the robotics API
2016-05-20 01:37:15 +00:00
case CMD_CAMERA_IMAGE_COMPLETED:
{
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
if (m_data->m_verboseOutput)
{
b3Printf("Camera image OK\n");
}
int numBytesPerPixel = 4;//RGBA
int numTotalPixels = serverCmd.m_sendPixelDataArguments.m_startingPixelIndex+
serverCmd.m_sendPixelDataArguments.m_numPixelsCopied+
serverCmd.m_sendPixelDataArguments.m_numRemainingPixels;
m_data->m_cachedCameraPixelsWidth = 0;
m_data->m_cachedCameraPixelsHeight = 0;
initial hookup of TinyRenderer to shared memory interface
2016-06-02 00:47:41 +00:00
int numPixels = serverCmd.m_sendPixelDataArguments.m_imageWidth*serverCmd.m_sendPixelDataArguments.m_imageHeight;
m_data->m_cachedCameraPixelsRGBA.reserve(numPixels*numBytesPerPixel);
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
m_data->m_cachedCameraDepthBuffer.resize(numTotalPixels);
add segmentation mask rendering to TinyRenderer and shared memory API similar to the zbuffer, but storing the object index (int) instead of float depth
2016-08-11 21:55:30 +00:00
m_data->m_cachedSegmentationMaskBuffer.resize(numTotalPixels);
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
m_data->m_cachedCameraPixelsRGBA.resize(numTotalPixels*numBytesPerPixel);
initial hookup of TinyRenderer to shared memory interface
2016-06-02 00:47:41 +00:00
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
unsigned char* rgbaPixelsReceived =
(unsigned char*)&m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[0];
implement first draft of pybullet.renderImage for synthetic camera remove a few debug printf from tinyrenderer
2016-06-07 01:54:05 +00:00
// printf("pixel = %d\n", rgbaPixelsReceived[0]);
initial hookup of TinyRenderer to shared memory interface
2016-06-02 00:47:41 +00:00
fix issues related to camera width/height add width,height as arguments to pybullet.renderImage(x,y,[viewMat4x4],[projMat4x4])
2016-06-07 23:11:58 +00:00
float* depthBuffer = (float*)&(m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[serverCmd.m_sendPixelDataArguments.m_numPixelsCopied*4]);
add segmentation mask rendering to TinyRenderer and shared memory API similar to the zbuffer, but storing the object index (int) instead of float depth
2016-08-11 21:55:30 +00:00
int* segmentationMaskBuffer = (int*)&(m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor[serverCmd.m_sendPixelDataArguments.m_numPixelsCopied*8]);
fix issues related to camera width/height add width,height as arguments to pybullet.renderImage(x,y,[viewMat4x4],[projMat4x4])
2016-06-07 23:11:58 +00:00
for (int i=0;i<serverCmd.m_sendPixelDataArguments.m_numPixelsCopied;i++)
{
m_data->m_cachedCameraDepthBuffer[i + serverCmd.m_sendPixelDataArguments.m_startingPixelIndex] = depthBuffer[i];
}
add segmentation mask rendering to TinyRenderer and shared memory API similar to the zbuffer, but storing the object index (int) instead of float depth
2016-08-11 21:55:30 +00:00
for (int i=0;i<serverCmd.m_sendPixelDataArguments.m_numPixelsCopied;i++)
{
m_data->m_cachedSegmentationMaskBuffer[i + serverCmd.m_sendPixelDataArguments.m_startingPixelIndex] = segmentationMaskBuffer[i];
}
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
for (int i=0;i<serverCmd.m_sendPixelDataArguments.m_numPixelsCopied*numBytesPerPixel;i++)
{
m_data->m_cachedCameraPixelsRGBA[i + serverCmd.m_sendPixelDataArguments.m_startingPixelIndex*numBytesPerPixel]
= rgbaPixelsReceived[i];
}
Expose a better API to allow any render engine to be used for the physics simulation when loading URDF/SDF files. See bullet3/examples/Importers/ImportURDFDemo/DefaultVisualShapeConverter.h Give the kuka_iiwa/model.urdf some blue color, not just orange, to mimick the original a bit better Preparation for the CMD_CAMERA_IMAGE_COMPLETED command, to expose a virtual camera to the robotics API
2016-05-20 01:37:15 +00:00
break;
}
case CMD_CAMERA_IMAGE_FAILED:
{
implement first draft of pybullet.renderImage for synthetic camera remove a few debug printf from tinyrenderer
2016-06-07 01:54:05 +00:00
b3Warning("Camera image FAILED\n");
Expose a better API to allow any render engine to be used for the physics simulation when loading URDF/SDF files. See bullet3/examples/Importers/ImportURDFDemo/DefaultVisualShapeConverter.h Give the kuka_iiwa/model.urdf some blue color, not just orange, to mimick the original a bit better Preparation for the CMD_CAMERA_IMAGE_COMPLETED command, to expose a virtual camera to the robotics API
2016-05-20 01:37:15 +00:00
break;
}
Add premake support to build pybullet, Windows and Linux tested, will enable Mac in next commit. Expose inverse dynamics to Bullet shared memory API, through b3CalculateInverseDynamicsCommandInit and b3GetStatusInverseDynamicsJointForces command/status. See PhysicsClientExeample or pybullet for usage. Add option for Windows and Linux to set python_lib_dir and python_include_dir for premake and --enable_pybullet option Expose inverse dynamics to pybullet: [force] = p.calculateInverseDynamics(objectIndex,[q],[qdot],[acc]) Thanks to Jeff Bingham for the suggestion.
2016-08-10 01:40:12 +00:00
case CMD_CALCULATED_INVERSE_DYNAMICS_COMPLETED:
{
break;
}
case CMD_CALCULATED_INVERSE_DYNAMICS_FAILED:
{
b3Warning("Inverse Dynamics computations failed");
break;
}
pybullet getOverlappingObjects added: report all object unique ids within a world space bounding box (aabb)
2016-11-10 19:22:22 +00:00
case CMD_REQUEST_AABB_OVERLAP_FAILED:
{
b3Warning("Overlapping object query failed");
break;
}
case CMD_REQUEST_AABB_OVERLAP_COMPLETED:
{
if (m_data->m_verboseOutput)
{
b3Printf("Overlapping object request completed");
}
int startOverlapIndex = serverCmd.m_sendOverlappingObjectsArgs.m_startingOverlappingObjectIndex;
int numOverlapCopied = serverCmd.m_sendOverlappingObjectsArgs.m_numOverlappingObjectsCopied;
m_data->m_cachedOverlappingObjects.resize(startOverlapIndex + numOverlapCopied);
b3OverlappingObject* objects = (b3OverlappingObject*)m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor;
for (int i = 0; i < numOverlapCopied; i++)
{
m_data->m_cachedOverlappingObjects[startOverlapIndex + i] = objects[i];
}
break;
}
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
case CMD_CONTACT_POINT_INFORMATION_COMPLETED:
{
if (m_data->m_verboseOutput)
{
b3Printf("Contact Point Information Request OK\n");
}
int startContactIndex = serverCmd.m_sendContactPointArgs.m_startingContactPointIndex;
int numContactsCopied = serverCmd.m_sendContactPointArgs.m_numContactPointsCopied;
m_data->m_cachedContactPoints.resize(startContactIndex+numContactsCopied);
b3ContactPointData* contactData = (b3ContactPointData*)m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor;
for (int i=0;i<numContactsCopied;i++)
{
m_data->m_cachedContactPoints[startContactIndex+i] = contactData[i];
}
break;
}
case CMD_CONTACT_POINT_INFORMATION_FAILED:
{
b3Warning("Contact Point Information Request failed");
break;
}
Add rudimentary 'saveWorld' command in shared memory API and pybullet, see examples/pybullet/saveWorld.py Use trilinear filtering instead of bilinear
2016-10-13 06:03:36 +00:00
Merge remote-tracking branch 'bp/master'
2016-10-13 06:07:41 +00:00
case CMD_SAVE_WORLD_COMPLETED:
break;
Add rudimentary 'saveWorld' command in shared memory API and pybullet, see examples/pybullet/saveWorld.py Use trilinear filtering instead of bilinear
2016-10-13 06:03:36 +00:00
case CMD_SAVE_WORLD_FAILED:
Merge remote-tracking branch 'bp/master'
2016-10-13 06:07:41 +00:00
{
b3Warning("Saving world failed");
break;
}
case CMD_CALCULATE_INVERSE_KINEMATICS_COMPLETED:
{
handle CMD_CALCULATE_INVERSE_KINEMATICS_COMPLETED/FAILED in PhysicsClientSharedMemory::processServerStatus
2016-10-12 17:46:07 +00:00
break;
}
case CMD_CALCULATE_INVERSE_KINEMATICS_FAILED:
{
b3Warning("Calculate Inverse Kinematics Request failed");
break;
}
first version of 'getVisualShapeData' to get visual shape information to allow external renderer with pybullet and shared memory robotics API b3InitRequestVisualShapeInformation/b3GetVisualShapeInformation in shared memory API
2016-10-19 05:05:28 +00:00
case CMD_VISUAL_SHAPE_INFO_COMPLETED:
{
if (m_data->m_verboseOutput)
{
b3Printf("Visual Shape Information Request OK\n");
}
int startVisualShapeIndex = serverCmd.m_sendVisualShapeArgs.m_startingVisualShapeIndex;
int numVisualShapesCopied = serverCmd.m_sendVisualShapeArgs.m_numVisualShapesCopied;
m_data->m_cachedVisualShapes.resize(startVisualShapeIndex + numVisualShapesCopied);
b3VisualShapeData* shapeData = (b3VisualShapeData*)m_data->m_testBlock1->m_bulletStreamDataServerToClientRefactor;
for (int i = 0; i < numVisualShapesCopied; i++)
{
m_data->m_cachedVisualShapes[startVisualShapeIndex + i] = shapeData[i];
}
Merge remote-tracking branch 'bp/master'
2016-10-13 06:07:41 +00:00
first version of 'getVisualShapeData' to get visual shape information to allow external renderer with pybullet and shared memory robotics API b3InitRequestVisualShapeInformation/b3GetVisualShapeInformation in shared memory API
2016-10-19 05:05:28 +00:00
break;
}
case CMD_VISUAL_SHAPE_INFO_FAILED:
{
b3Warning("Visual Shape Info Request failed");
break;
}
Add programmatic render API and a basic test.
2016-10-20 17:56:44 +00:00
case CMD_VISUAL_SHAPE_UPDATE_COMPLETED:
{
break;
}
case CMD_VISUAL_SHAPE_UPDATE_FAILED:
{
b3Warning("Visual Shape Update failed");
break;
}
Add loading texture API.
2016-10-21 06:40:30 +00:00
case CMD_LOAD_TEXTURE_COMPLETED:
{
break;
}
case CMD_LOAD_TEXTURE_FAILED:
{
b3Warning("Load texture failed");
break;
}
add pybullet loadBullet, saveBullet (work-in-progress) and placeholder for loadMJCF.
2016-11-12 02:07:42 +00:00
case CMD_BULLET_LOADING_COMPLETED:
{
break;
}
case CMD_BULLET_LOADING_FAILED:
{
b3Warning("Load .bullet failed");
break;
}
case CMD_BULLET_SAVING_FAILED:
{
b3Warning("Save .bullet failed");
break;
}
case CMD_MJCF_LOADING_FAILED:
{
b3Warning("Load .mjcf failed");
break;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
default: {
handle CMD_CALCULATE_INVERSE_KINEMATICS_COMPLETED/FAILED in PhysicsClientSharedMemory::processServerStatus
2016-10-12 17:46:07 +00:00
b3Error("Unknown server status %d\n", serverCmd.m_type);
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
btAssert(0);
}
};
m_data->m_testBlock1->m_numProcessedServerCommands++;
// we don't have more than 1 command outstanding (in total, either server or client)
btAssert(m_data->m_testBlock1->m_numProcessedServerCommands ==
m_data->m_testBlock1->m_numServerCommands);
if (m_data->m_testBlock1->m_numServerCommands ==
m_data->m_testBlock1->m_numProcessedServerCommands) {
m_data->m_waitingForServer = false;
} else {
m_data->m_waitingForServer = true;
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
}
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
if (serverCmd.m_type == CMD_SDF_LOADING_COMPLETED)
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
{
int numBodies = serverCmd.m_sdfLoadedArgs.m_numBodies;
if (numBodies>0)
{
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
m_data->m_tempBackupServerStatus = m_data->m_lastServerStatus;
for (int i=0;i<numBodies;i++)
{
m_data->m_bodyIdsRequestInfo.push_back(serverCmd.m_sdfLoadedArgs.m_bodyUniqueIds[i]);
}
int bodyId = m_data->m_bodyIdsRequestInfo[m_data->m_bodyIdsRequestInfo.size()-1];
m_data->m_bodyIdsRequestInfo.pop_back();
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
//now transfer the information of the individual objects etc.
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
command.m_type = CMD_REQUEST_BODY_INFO;
command.m_sdfRequestInfoArgs.m_bodyUniqueId = bodyId;
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
submitClientCommand(command);
return 0;
}
}
implement getJointInfo for objects loaded through SDF
2016-06-16 01:01:41 +00:00
if (serverCmd.m_type == CMD_BODY_INFO_COMPLETED)
{
//are there any bodies left to be processed?
if (m_data->m_bodyIdsRequestInfo.size())
{
int bodyId = m_data->m_bodyIdsRequestInfo[m_data->m_bodyIdsRequestInfo.size()-1];
m_data->m_bodyIdsRequestInfo.pop_back();
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
//now transfer the information of the individual objects etc.
command.m_type = CMD_REQUEST_BODY_INFO;
command.m_sdfRequestInfoArgs.m_bodyUniqueId = bodyId;
submitClientCommand(command);
return 0;
} else
{
m_data->m_lastServerStatus = m_data->m_tempBackupServerStatus;
}
}
pybullet getOverlappingObjects added: report all object unique ids within a world space bounding box (aabb)
2016-11-10 19:22:22 +00:00
if (serverCmd.m_type == CMD_REQUEST_AABB_OVERLAP_COMPLETED)
{
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
if (serverCmd.m_sendOverlappingObjectsArgs.m_numRemainingOverlappingObjects > 0 && serverCmd.m_sendOverlappingObjectsArgs.m_numOverlappingObjectsCopied)
{
command.m_type = CMD_REQUEST_AABB_OVERLAP;
command.m_requestOverlappingObjectsArgs.m_startingOverlappingObjectIndex = serverCmd.m_sendOverlappingObjectsArgs.m_startingOverlappingObjectIndex + serverCmd.m_sendOverlappingObjectsArgs.m_numOverlappingObjectsCopied;
submitClientCommand(command);
return 0;
}
}
Fix pybullet Windows build errors: C99 requires variables to be defined at the start of the function. Improve CMake Windows support to build PyBullet (BUILD_PYBULLET) Implement b3LoadSdfCommandInit in shared memory API Implement pybullet SDF loading binding, in loadSDF API TODO for SDF support is provide way to query object/link/joint information.
2016-06-13 17:11:28 +00:00
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
if (serverCmd.m_type == CMD_CONTACT_POINT_INFORMATION_COMPLETED)
{
implement pybullet.getContactPointData(), two optional object unique ids as filter returns a pylist of contact points. Each point has the following data: 0 int m_contactFlags;//unused for now 1 int m_bodyUniqueIdA; 2 int m_bodyUniqueIdB; 3 int m_linkIndexA; 4 int m_linkIndexB; 5-6-7 double m_positionOnAInWS[3];//contact point location on object A, in world space coordinates 8-9-10 double m_positionOnBInWS[3];//contact point location on object A, in world space coordinates 11-12-13 double m_contactNormalOnBInWS[3];//the separating contact normal, pointing from object B towards object A 14 double m_contactDistance;//negative number is penetration, positive is distance. 15 double m_normalForce;
2016-09-02 01:28:39 +00:00
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
if (serverCmd.m_sendContactPointArgs.m_numRemainingContactPoints>0 && serverCmd.m_sendContactPointArgs.m_numContactPointsCopied)
{
command.m_type = CMD_REQUEST_CONTACT_POINT_INFORMATION;
command.m_requestContactPointArguments.m_startingContactPointIndex = serverCmd.m_sendContactPointArgs.m_startingContactPointIndex+serverCmd.m_sendContactPointArgs.m_numContactPointsCopied;
command.m_requestContactPointArguments.m_objectAIndexFilter = -1;
command.m_requestContactPointArguments.m_objectBIndexFilter = -1;
submitClientCommand(command);
return 0;
}
}
first version of 'getVisualShapeData' to get visual shape information to allow external renderer with pybullet and shared memory robotics API b3InitRequestVisualShapeInformation/b3GetVisualShapeInformation in shared memory API
2016-10-19 05:05:28 +00:00
if (serverCmd.m_type == CMD_VISUAL_SHAPE_INFO_COMPLETED)
{
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
if (serverCmd.m_sendVisualShapeArgs.m_numRemainingVisualShapes >0 && serverCmd.m_sendVisualShapeArgs.m_numVisualShapesCopied)
{
command.m_type = CMD_REQUEST_VISUAL_SHAPE_INFO;
command.m_requestVisualShapeDataArguments.m_startingVisualShapeIndex = serverCmd.m_sendVisualShapeArgs.m_startingVisualShapeIndex + serverCmd.m_sendVisualShapeArgs.m_numVisualShapesCopied;
command.m_requestVisualShapeDataArguments.m_bodyUniqueId = serverCmd.m_sendVisualShapeArgs.m_bodyUniqueId;
submitClientCommand(command);
return 0;
}
}
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
if (serverCmd.m_type == CMD_CAMERA_IMAGE_COMPLETED)
{
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
add pybullet render API with yaw/pitch/roll option add testrender.py file allow option to enable OpenGL hardware renderer in multithreaded sim b3RequestCameraImageSelectRenderer(commandHandle,ER_BULLET_HARDWARE_OPENGL);
2016-08-08 21:23:44 +00:00
if (serverCmd.m_sendPixelDataArguments.m_numRemainingPixels > 0 && serverCmd.m_sendPixelDataArguments.m_numPixelsCopied)
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
{
// continue requesting remaining pixels
command.m_type = CMD_REQUEST_CAMERA_IMAGE_DATA;
command.m_requestPixelDataArguments.m_startPixelIndex =
serverCmd.m_sendPixelDataArguments.m_startingPixelIndex +
serverCmd.m_sendPixelDataArguments.m_numPixelsCopied;
submitClientCommand(command);
return 0;
} else
{
m_data->m_cachedCameraPixelsWidth = serverCmd.m_sendPixelDataArguments.m_imageWidth;
m_data->m_cachedCameraPixelsHeight = serverCmd.m_sendPixelDataArguments.m_imageHeight;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
}
if ((serverCmd.m_type == CMD_DEBUG_LINES_COMPLETED) &&
(serverCmd.m_sendDebugLinesArgs.m_numRemainingDebugLines > 0)) {
SharedMemoryCommand& command = m_data->m_testBlock1->m_clientCommands[0];
// continue requesting debug lines for drawing
command.m_type = CMD_REQUEST_DEBUG_LINES;
command.m_requestDebugLinesArguments.m_startingLineIndex =
serverCmd.m_sendDebugLinesArgs.m_numDebugLines +
serverCmd.m_sendDebugLinesArgs.m_startingLineIndex;
submitClientCommand(command);
return 0;
}
return &m_data->m_lastServerStatus;
} else {
if (m_data->m_verboseOutput) {
b3Printf("m_numServerStatus = %d, processed = %d\n",
m_data->m_testBlock1->m_numServerCommands,
m_data->m_testBlock1->m_numProcessedServerCommands);
}
}
return 0;
}
bool PhysicsClientSharedMemory::canSubmitCommand() const {
return (m_data->m_isConnected && !m_data->m_waitingForServer);
}
struct SharedMemoryCommand* PhysicsClientSharedMemory::getAvailableSharedMemoryCommand() {
Physics runs in a separate thread from rendering in PhysicsServerExample (preliminary) Improve rendering performance. OpenVR experience is smooth now. commit needs a bit more testing before pushing in main repo.
2016-07-08 02:24:44 +00:00
static int sequence = 0;
m_data->m_testBlock1->m_clientCommands[0].m_sequenceNumber = sequence++;
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
return &m_data->m_testBlock1->m_clientCommands[0];
}
bool PhysicsClientSharedMemory::submitClientCommand(const SharedMemoryCommand& command) {
/// at the moment we allow a maximum of 1 outstanding command, so we check for this
// once the server processed the command and returns a status, we clear the flag
// "m_data->m_waitingForServer" and allow submitting the next command
if (!m_data->m_waitingForServer) {
if (&m_data->m_testBlock1->m_clientCommands[0] != &command) {
m_data->m_testBlock1->m_clientCommands[0] = command;
}
m_data->m_testBlock1->m_numClientCommands++;
m_data->m_waitingForServer = true;
return true;
}
return false;
}
void PhysicsClientSharedMemory::uploadBulletFileToSharedMemory(const char* data, int len) {
btAssert(len < SHARED_MEMORY_MAX_STREAM_CHUNK_SIZE);
if (len >= SHARED_MEMORY_MAX_STREAM_CHUNK_SIZE) {
b3Warning("uploadBulletFileToSharedMemory %d exceeds max size %d\n", len,
SHARED_MEMORY_MAX_STREAM_CHUNK_SIZE);
} else {
for (int i = 0; i < len; i++) {
m_data->m_testBlock1->m_bulletStreamDataClientToServer[i] = data[i];
}
}
}
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
void PhysicsClientSharedMemory::getCachedCameraImage(struct b3CameraImageData* cameraData)
{
cameraData->m_pixelWidth = m_data->m_cachedCameraPixelsWidth;
cameraData->m_pixelHeight = m_data->m_cachedCameraPixelsHeight;
revert accidently renaming of m_depthvalues4
2016-08-12 20:55:37 +00:00
cameraData->m_depthValues = m_data->m_cachedCameraDepthBuffer.size() ? &m_data->m_cachedCameraDepthBuffer[0] : 0;
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
cameraData->m_rgbColorData = m_data->m_cachedCameraPixelsRGBA.size() ? &m_data->m_cachedCameraPixelsRGBA[0] : 0;
add segmentation mask rendering to TinyRenderer and shared memory API similar to the zbuffer, but storing the object index (int) instead of float depth
2016-08-11 21:55:30 +00:00
cameraData->m_segmentationMaskValues = m_data->m_cachedSegmentationMaskBuffer.size()?&m_data->m_cachedSegmentationMaskBuffer[0] : 0;
preparation to receive camera image data from physics server increase shadowmap world size default to 50 units (meter), 10 units (meter) was too small for most examples.
2016-05-31 17:23:04 +00:00
}
add first draft of contact point query in shared memory API b3SharedMemoryCommandHandle b3InitRequestContactPointInformation(b3PhysicsClientHandle physClient); void b3SetContactFilterBodyA(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdA); void b3SetContactFilterBodyB(b3SharedMemoryCommandHandle commandHandle, int bodyUniqueIdB); void b3GetContactPointInformation(b3PhysicsClientHandle physClient, struct b3ContactInformation* contactPointData); Implemented for PhysicsClientSharedMemory, not for PhysicsDirect yet. Add btCollisionObject::setUserIndex2
2016-09-01 20:30:07 +00:00
void PhysicsClientSharedMemory::getCachedContactPointInformation(struct b3ContactInformation* contactPointData)
{
contactPointData->m_numContactPoints = m_data->m_cachedContactPoints.size();
contactPointData->m_contactPointData = contactPointData->m_numContactPoints? &m_data->m_cachedContactPoints[0] : 0;
}
pybullet getOverlappingObjects added: report all object unique ids within a world space bounding box (aabb)
2016-11-10 19:22:22 +00:00
void PhysicsClientSharedMemory::getCachedOverlappingObjects(struct b3AABBOverlapData* overlappingObjects)
{
overlappingObjects->m_numOverlappingObjects = m_data->m_cachedOverlappingObjects.size();
overlappingObjects->m_overlappingObjects = m_data->m_cachedOverlappingObjects.size() ?
&m_data->m_cachedOverlappingObjects[0] : 0;
}
first version of 'getVisualShapeData' to get visual shape information to allow external renderer with pybullet and shared memory robotics API b3InitRequestVisualShapeInformation/b3GetVisualShapeInformation in shared memory API
2016-10-19 05:05:28 +00:00
void PhysicsClientSharedMemory::getCachedVisualShapeInformation(struct b3VisualShapeInformation* visualShapesInfo)
{
visualShapesInfo->m_numVisualShapes = m_data->m_cachedVisualShapes.size();
visualShapesInfo->m_visualShapeData = visualShapesInfo->m_numVisualShapes ? &m_data->m_cachedVisualShapes[0] : 0;
}
make PhysicsClient an interface class, and move shared memory implementation of physics client into PhysicsClientSharedMemory class/file. This will allow implementation of UDP and other transport protocols to a physics server.
2015-09-25 05:50:34 +00:00
const float* PhysicsClientSharedMemory::getDebugLinesFrom() const {
if (m_data->m_debugLinesFrom.size()) {
return &m_data->m_debugLinesFrom[0].m_x;
}
return 0;
}
const float* PhysicsClientSharedMemory::getDebugLinesTo() const {
if (m_data->m_debugLinesTo.size()) {
return &m_data->m_debugLinesTo[0].m_x;
}
return 0;
}
const float* PhysicsClientSharedMemory::getDebugLinesColor() const {
if (m_data->m_debugLinesColor.size()) {
return &m_data->m_debugLinesColor[0].m_x;
}
return 0;
}
int PhysicsClientSharedMemory::getNumDebugLines() const { return m_data->m_debugLinesFrom.size(); }
Reference in New Issue Copy Permalink