Good and Bad of JT

• + Cheaper evaluation step than pseudoinverse

• + No singularities

• - Scaling problems

  • J+ has nice property that solution has minimal norm at every step.
  • JT doesn't have this property. Joints far from end effector experience larger torques, hence take disproportionately large steps.
  • Can throw in a constant diagonal scaling matrix to counteract some scaling probs

• - Slower to converge than J+

  • (2x slower according to Das, Slotine & Sheridan)

Notes: