<link name="dummy">      <inertial>      <mass value="0"/>      <inertia ixx="1.0" ixy="0.0" ixz="0.0" iyy="1.0" iyz="0.0" izz="1.0"/>    </inertial>  </link> <joint name="base_to_dummy" type="fixed">    <parent link="planeLink"/>    <child link="dummy"/>    <origin xyz="0 0 0.1"/>  </joint>

import pybullet as pp.connect(p.GUI)plane = p.loadURDF("plane.urdf")dummyLink=0s = p.getLinkState(plane,dummyLink, computeForwardKinematics=True, computeLinkVelocity=True)pos = s[0]orn = s[1]linvel = s[6]angvel = s[7]print("world pos=",pos)

Statistics: Posted by Erwin Coumans — Mon Oct 15, 2018 9:12 pm

It is better to add some fixed dummy links in the URDF file. Then let getLinkState compute the world space position and world space velocity for those dummy links.

You could create a small pybullet test and share it on github, plot the curve with matplotlib (keep all dependencies to a minimum) link to it, so someone can have a look at it.

You should be able to model robots and cars pretty accurate with PyBullet.

Statistics: Posted by mimeakadug — Mon Oct 15, 2018 3:58 pm

Your story is a bit complicated to me and maybe some code would clarify. I have experienced difficulties myself with torque driven motors. I tried a few work arounds, played with various friction components, but still my torque motor did not meet perfectly with analytical calculations. If you do not really need the torque-motor with correct mass-acceleration etc, then you can simply set rotation velocity or angle of the wheel at every time-step following your analytical calculation. Below is some code using a torque motor to swing an inertial mass. It is not perfect and you see a lot of commented commands I tried to play with. It may just help you a small step further. BTW if you find better ways I'm interested too.

Code:

import pybullet as pimport timeimport numpy as npp.connect(p.GUI)p.createCollisionShape(p.GEOM_PLANE)p.createMultiBody(0,0)#topboxHalfLength = 0.05boxHalfWidth = 0.05boxHalfHeight = 0.5body = p.createCollisionShape(p.GEOM_BOX,halfExtents=[boxHalfLength,boxHalfWidth,boxHalfHeight])pin = p.createCollisionShape(p.GEOM_BOX,halfExtents=[0.05,0.05,0.05])wgt = p.createCollisionShape(p.GEOM_BOX,halfExtents=[0.05,0.05,0.05])mass = 10000visualShapeId = -1nlnk=2link_Masses=[0,1]linkCollisionShapeIndices=[pin,wgt]linkVisualShapeIndices=[-1]*nlnklinkPositions=[[0.0, 0.0, 0.0],[1.0,0,0]]linkOrientations=[[0,0,0,1]]*nlnklinkInertialFramePositions=[[0,0,0]]*nlnklinkInertialFrameOrientations=[[0,0,0,1]]*nlnkindices=[0,1]jointTypes=[p.JOINT_REVOLUTE,p.JOINT_REVOLUTE]axis=[[0,0,1],[0,1,0]]basePosition = [0,0,0.5]baseOrientation = [0,0,0,1]block = p.createMultiBody(mass,body,visualShapeId,basePosition,baseOrientation,                        linkMasses=link_Masses,                        linkCollisionShapeIndices=linkCollisionShapeIndices,                        linkVisualShapeIndices=linkVisualShapeIndices,                        linkPositions=linkPositions,                        linkOrientations=linkOrientations,                        linkInertialFramePositions=linkInertialFramePositions,                        linkInertialFrameOrientations=linkInertialFrameOrientations,                        linkParentIndices=indices,                        linkJointTypes=jointTypes,                        linkJointAxis=axis)p.setGravity(0,0,-0.0)p.setTimeStep(0.01)p.setRealTimeSimulation(0)#p.resetBaseVelocity(top,angularVelocity=[0,0,30])p.changeDynamics(block,0,lateralFriction=0,angularDamping=0.000,linearDamping=0.000,spinningFriction=0,rollingFriction=0)p.changeDynamics(block,1,lateralFriction=0,angularDamping=0.000,linearDamping=0.000,spinningFriction=0,rollingFriction=0)#p.changeDynamics(block,-1,lateralFriction=10000,angularDamping=1.000,linearDamping=0.000,spinningFriction=10000)p.resetDebugVisualizerCamera( cameraDistance=2, cameraYaw=10, cameraPitch=-20, cameraTargetPosition=[0.0, 0.0, 0.25])t0=time.time()t=time.time()while ((t-t0)<3):    t=time.time()p.resetBasePositionAndOrientation(block,[0,0,0.5],[0,0,0,1])p.enableJointForceTorqueSensor(block,0,enableSensor=1)#p.setJointMotorControl2(block,0,p.POSITION_CONTROL,targetPosition=-1,force=1000,maxVelocity=1)p.setJointMotorControl2(block,0,p.VELOCITY_CONTROL,0,0,0)#p.setJointMotorControl2(block,0,controlMode=p.TORQUE_CONTROL,force=-20)#dum2=p.getDynamicsInfo(block,0)dum2=p.getJointState(block,0)#init robot positiont0=time.time()t=time.time()its=0while ((t-t0)<5):    its+=1    #p.enableJointForceTorqueSensor(block,0,enableSensor=1)    p.setJointMotorControl2(block,0,controlMode=p.TORQUE_CONTROL,force=-1)    dum2=p.getJointState(block,0)    t=time.time()        dum=p.getLinkState(block,1)    xft=dum[0]    dum=p.rayTest(xft-np.array([0,0,0.051]),xft-np.array([0,0,1.051]))    #print(t-t0,1*dum[0][2])    p.stepSimulation()    time.sleep(1./100.)    #print(t-t0,80*dum[0][2])p.disconnect()

Statistics: Posted by mimeakadug — Mon Oct 15, 2018 3:56 pm

Statistics: Posted by Erwin Coumans — Fri Oct 05, 2018 1:30 am

import pybullet as pimport timeimport numpy as npp.connect(p.GUI)p.createCollisionShape(p.GEOM_PLANE)p.createMultiBody(0,0)#topboxHalfLength = 0.05boxHalfWidth = 0.05boxHalfHeight = 0.5body = p.createCollisionShape(p.GEOM_BOX,halfExtents=[boxHalfLength,boxHalfWidth,boxHalfHeight])pin = p.createCollisionShape(p.GEOM_BOX,halfExtents=[0.05,0.05,0.05])wgt = p.createCollisionShape(p.GEOM_BOX,halfExtents=[0.05,0.05,0.05])mass = 10000visualShapeId = -1nlnk=2link_Masses=[0,1]linkCollisionShapeIndices=[pin,wgt]linkVisualShapeIndices=[-1]*nlnklinkPositions=[[0.0, 0.0, 0.0],[1.0,0,0]]linkOrientations=[[0,0,0,1]]*nlnklinkInertialFramePositions=[[0,0,0]]*nlnklinkInertialFrameOrientations=[[0,0,0,1]]*nlnkindices=[0,1]jointTypes=[p.JOINT_REVOLUTE,p.JOINT_REVOLUTE]axis=[[0,0,1],[0,1,0]]basePosition = [0,0,0.5]baseOrientation = [0,0,0,1]block = p.createMultiBody(mass,body,visualShapeId,basePosition,baseOrientation,                        linkMasses=link_Masses,                        linkCollisionShapeIndices=linkCollisionShapeIndices,                        linkVisualShapeIndices=linkVisualShapeIndices,                        linkPositions=linkPositions,                        linkOrientations=linkOrientations,                        linkInertialFramePositions=linkInertialFramePositions,                        linkInertialFrameOrientations=linkInertialFrameOrientations,                        linkParentIndices=indices,                        linkJointTypes=jointTypes,                        linkJointAxis=axis)p.setGravity(0,0,-0.0)p.setTimeStep(0.01)p.setRealTimeSimulation(0)#p.resetBaseVelocity(top,angularVelocity=[0,0,30])p.changeDynamics(block,0,lateralFriction=0,angularDamping=0.000,linearDamping=0.000,spinningFriction=0,rollingFriction=0)p.changeDynamics(block,1,lateralFriction=0,angularDamping=0.000,linearDamping=0.000,spinningFriction=0,rollingFriction=0)#p.changeDynamics(block,-1,lateralFriction=10000,angularDamping=1.000,linearDamping=0.000,spinningFriction=10000)p.resetDebugVisualizerCamera( cameraDistance=2, cameraYaw=10, cameraPitch=-20, cameraTargetPosition=[0.0, 0.0, 0.25])t0=time.time()t=time.time()while ((t-t0)<3):    t=time.time()p.resetBasePositionAndOrientation(block,[0,0,0.5],[0,0,0,1])p.enableJointForceTorqueSensor(block,0,enableSensor=1)#p.setJointMotorControl2(block,0,p.POSITION_CONTROL,targetPosition=-1,force=1000,maxVelocity=1)p.setJointMotorControl2(block,0,p.VELOCITY_CONTROL,0,0,0)#p.setJointMotorControl2(block,0,controlMode=p.TORQUE_CONTROL,force=-20)#dum2=p.getDynamicsInfo(block,0)dum2=p.getJointState(block,0)#init robot positiont0=time.time()t=time.time()its=0while ((t-t0)<5):    its+=1    #p.enableJointForceTorqueSensor(block,0,enableSensor=1)    p.setJointMotorControl2(block,0,controlMode=p.TORQUE_CONTROL,force=-1)    dum2=p.getJointState(block,0)    t=time.time()        dum=p.getLinkState(block,1)    xft=dum[0]    dum=p.rayTest(xft-np.array([0,0,0.051]),xft-np.array([0,0,1.051]))    #print(t-t0,1*dum[0][2])    p.stepSimulation()    time.sleep(1./100.)    #print(t-t0,80*dum[0][2])p.disconnect()

Statistics: Posted by richardbloemenkamp — Fri Sep 28, 2018 5:15 pm

,1,1)[4] I get the position of the axle ends which is also the origin of my inertial coordinate system
. Using getLinkState(car,track,1,1)[5] I get the orientation of the axle ends coordinate system.
9). I move the points on the wheel to the origin using 7). and rotate them using . and again using 7). I translate it back. This way I get the world coordinates of the points on the wheels.

This approach however is not giving me expected plots of position and velocity. For points on the wheel rim, they should follow a cycloid (https://en.wikipedia.org/wiki/Cycloid). The speed should change from 0 to 2v m/s where v is the translational velocity of the the car as it moves. However Neither of these plots are as expected. Does anyone have a simpler approach?

thanks for your time!

Statistics: Posted by mimeakadug — Mon Sep 24, 2018 3:37 pm