Skip to content
Tutorials Center
My learning

Topics

Lesson 8 of 38 ยท Kinematics

Jacobian and Velocity Kinematics

Relate joint speeds to tip speed with the Jacobian matrix, find singularities, measure manipulability, and map tip forces to joint torques using Python.

Forward and inverse kinematics deal with positions. Robots also need velocities: how fast must each joint turn so the tip moves along a line at 0.1 m/s? The answer is the Jacobian, a matrix of partial derivatives that turns joint velocities into tip velocity. It is one of the most useful objects in robotics: it also explains singularities, powers numerical inverse kinematics and links tip forces to motor torques.

From positions to velocities

Forward kinematics gives the tip position as a function of the joints: x = f(q). Differentiating with respect to time with the chain rule gives

แบ‹ = J(q) qฬ‡

where qฬ‡ is the vector of joint velocities and J = โˆ‚f/โˆ‚q is the Jacobian. Each column says how the tip moves when only that joint moves. Because it depends on q, the Jacobian changes as the arm moves. For a planar two-link arm:

Text
J = [[-l1 sin ฮธ1 - l2 sin(ฮธ1+ฮธ2),  -l2 sin(ฮธ1+ฮธ2)],
     [ l1 cos ฮธ1 + l2 cos(ฮธ1+ฮธ2),   l2 cos(ฮธ1+ฮธ2)]]

Column 1 is the tip velocity per unit shoulder speed; the shoulder rotates the whole arm, so its lever arm is the full distance to the tip. Column 2 involves only the second link.

You can verify any Jacobian numerically by nudging each joint a tiny amount and watching the tip.

Python Analytic Jacobian versus finite differences

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Resolved-rate motion

To make the tip move with a desired velocity v, invert the relationship: qฬ‡ = Jโปยน v (or the pseudo-inverse Jโบ v if J is not square). Integrating the joint velocities over small time steps dt makes the tip follow a straight line without solving IK for each point. This is called resolved-rate control.

Python Drawing a straight line with resolved-rate control

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

A little drift remains because each step uses a first-order approximation; closing the loop with feedback on the tip error reduces it.

Singularities

If the Jacobian loses rank, det(J) = 0, the arm is at a singularity. In that pose some tip velocity directions are impossible, and reaching nearby directions demands very large joint speeds. For the two-link arm,

det(J) = l1 l2 sin ฮธ2

so singularities occur when ฮธ2 = 0 (arm fully stretched) or ฮธ2 = 180ยฐ (folded back). At the edge of the workspace the tip cannot move outward no matter how fast the joints turn.

A smoother measure is manipulability, w = sqrt(det(J Jแต€)); it is large when the arm moves easily in all directions and drops to zero at singularities. The singular values of J tell the speed gain in each direction; the ratio of largest to smallest (the condition number) says how uneven the motion is.

Python Manipulability as the elbow straightens

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Planners often steer away from poses with low manipulability so the arm keeps its agility.

Forces and torques: the Jacobian transpose

By the principle of virtual work, the same Jacobian relates forces to torques:

ฯ„ = Jแต€ F

where F is the force at the tip and ฯ„ the vector of joint torques needed to hold it. This explains why a stretched-out arm needs huge shoulder torque to hold a weight (long lever arm), and why near a singularity a small tip force can load the joints very differently from what you expect. Motor sizing starts from this relation, which links back to Motors and Actuators.

Python Torque needed to hold a 2 kg load

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

(Link weight is ignored here to keep the idea clear; a real design adds it.)

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What does the Jacobian relate?

    x-dot = J(q) q-dot: it maps joint velocities to tip velocity.
  2. 2. For the planar two-link arm, when does the Jacobian become singular?

    det(J) = l1 l2 sin(theta2), which is zero for a fully stretched or folded arm.
  3. 3. What does tau = J^T F tell you?

    Virtual work links tip forces to joint torques through the transposed Jacobian.
  4. 4. Why is resolved-rate control not exact?

    Integrating small steps of a first-order model drifts; adding feedback on the tip error fixes it.

Summary

  • The Jacobian maps joint velocities to tip velocity and depends on the current pose.
  • Resolved-rate control inverts it to move the tip along a desired velocity.
  • Singularities (rank loss) are where the arm cannot move in some direction; manipulability measures distance from them.
  • The transpose maps tip forces to joint torques, which sizes the motors.

Next: Mobile Robot Kinematics