Simulating an KUKA LBR iwaa

[Decker]

Hey, there!

Currently I am attempting to simulate a KUKA LBR iwaa robot using Bullet3, an in-house renderer and C++.
I tried two approaches:

• - btHingeConstraint
  - btMultibody

Both approaches failed. The former approach fails the following way:
All links of the robot 'explode' into different directions. At this point, I am not sure what I am doing wrong.
Given an URDF description, how could one calculate the proper position of the first joint?

Using the latter approach, my application crashes on the first simulation step, telling me that one joint is untyped,
although, after verifying with the Debugger, every joint created is an revolute-joint.

Is there something I am doing wrong?

NOTE: I cannot rely on BulletRobotics, thus I am parsing the URDF description and creating the render objects myself.

[steven]

Could you provide a complete simple test? thanks.

[Decker]

Hey there!
Sorry for the late reply; I was been busy with this semesters start.

Today I found time to tinker around with it again and settled for hinge constraints. My goal was to get just one hinge to work.
Let's suppose we have the following URDF definition (just an excerpt from bullet's kuka.urdf):

Code: Select all

  <link name="calib_kuka_arm_base_link">
    <inertial>
      <mass value="0"/>
      <!-- static base, disable dynamics for this link -->
      <origin rpy="0 0 0" xyz="0 0 0.055"/>
      <inertia ixx="0.00381666666667" ixy="0" ixz="0" iyy="0.0036" iyz="0" izz="0.00381666666667"/>
    </inertial>
    <visual>
      <origin rpy="0 0 0" xyz="0 0 0"/>
      <geometry>
        <mesh filename="assets/models/kuka_lwr/arm_base.ply"/>
      </geometry>
      <material name="Orange"/>
    </visual>
    <collision>
      <origin rpy="0 0 0" xyz="0 0 0"/>
      <geometry>
        <mesh filename="assets/models/kuka_lwr/arm_base_convex.ply"/>
      </geometry>
    </collision>
  </link>
  <joint name="kuka_arm_0_joint" type="revolute">
    <origin rpy="0 0 0" xyz="0 0 0.11"/>
    <axis xyz="0 0 1"/>
    <limit effort="204" lower="-2.96705972839" upper="2.96705972839" velocity="1.91986217719"/>
    <dynamics damping="0.1"/>
    <parent link="calib_kuka_arm_base_link"/>
    <child link="kuka_arm_1_link"/>
  </joint>
  <link name="kuka_arm_1_link">
    <inertial>
      <mass value="2.0"/>
      <origin rpy="0 0 0" xyz="0 0.04 0.130"/>
      <inertia ixx="0.0136666666667" ixy="0" ixz="0" iyy="0.0118666666667" iyz="0" izz="0.003"/>
    </inertial>
    <visual>
      <origin rpy="0 0 3.14159265359" xyz="0 0 0"/>
      <geometry>
        <mesh filename="assets/models/kuka_lwr/arm_segment_a.ply"/>
      </geometry>
      <material name="Orange"/>
    </visual>
    <collision>
      <origin rpy="0 0 3.14159265359" xyz="0 0 0"/>
      <geometry>
        <mesh filename="assets/models/kuka_lwr/arm_segment_a_convex.ply"/>
      </geometry>
    </collision>
  </link>

The way I understood the URDF specification, the pivot in a would be (0, 0, 0.11), pivot in b would be(0, 0, 0) and the axis in a and b (0, 0, 1).
So I assumed, that following code would work:

Code: Select all

   for (std::size_t it = 1; it < link_count; ++it) {
      const auto& link = _robot->link(it);
      const auto& joint = link.parent_joint();

      btVector3   pivot_in_b = joint.origin().getOrigin(); // Specifies the offset off the joint in its local frame; here (0, 0, 0.11)
      btVector3   pivot_in_a = { 0, 0, 0 };                        // Is this assumption right?

      // TODO: Offset collision shape???

      auto hinge = new btHingeConstraint(
         *_rigid_bodies[it - 1],
         *_rigid_bodies[it],
         pivot_in_a,
         pivot_in_b,
         joint.axis(),
         joint.axis()
      );

      hinge->enableMotor(true);
      hinge->setMaxMotorImpulse(1000);

      auto component = new constraint_component{};
      component->constraint = hinge;
      _entities[it].add_component(component);
   }

Is this the right assumption? If not, how would I calculate the pivot points and axis?