In the previous video, we derived these equations of motion for a rigid body, for wrenches,

twists, and accelerations defined in a body frame {b}.

In this video, we express these equations in a form analogous to the equations of motion

for a body that only rotates.

To do this, we define the 6-by-6 spatial inertia matrix G_b.

The top left 3-by-3 submatrix is the inertia matrix I_b, and the bottom right 3-by-3 submatrix

is a diagonal matrix, the mass of the body multiplied by the 3-by-3 identity matrix.

With this, the kinetic energy of a rotating and translating rigid body can be written

as the sum of the rotational kinetic energy and the linear kinetic energy, or simply one-half

times the body twist transpose times G_b times the body twist.

Also, just like the mass matrix of a robot and the rotational inertia matrix I_b, the

spatial inertia matrix G_b is symmetric and positive definite.

Before proceeding further, we need to define an operation on 6-dimensional twists that

is analogous to the cross-product operation on 3-dimensional vectors.

Remember that we use the bracket notation to write the cross product of omega_1 and

omega_2 in R3.

The result is a 3-vector.

The 3-by-3 little so(3) matrix form of this cross product is bracket-omega_1 times bracket-omega_2

minus bracket-omega_2 times bracket-omega_1.

For 6-dimensional twists, our analogy to the cross product is bracket-V_1 times bracket-V_2

minus bracket-V_2 times bracket-V_1, a 4-by-4 matrix in little se(3).

In vector form, this is little-adV_1 times V_2, where the little adjoint of a twist V

is defined by the 6-by-6 matrix shown here.

Little-adV_1 times V_2 is called the Lie bracket of V_1 and V_2.

The Lie bracket of V_1 and V_2 is an acceleration, measuring how motion along the twist V_2 would

change if the body follows the twist V_1.

The matrix form of the Lie bracket is analogous to the matrix form of the cross product.

Returning now to our equations of motion, after a little manipulation we find that these

equations can be expressed as this 6-vector equation: the wrench F_b equals G_b times

V_b-dot minus little-adV_b-transpose times G_b times V_b.

Notice that the second term is a velocity-product term.

This equation is analogous to the equation for a rotating rigid body, replacing the rotational

inertia matrix I_b with the spatial inertia matrix G_b, the angular velocity omega_b with

the twist V_b, and the cross product with omega_b with a Lie bracket with V_b.

Now that we have the equation of motion in the {b} frame, we could ask what the equation

of motion is in a different frame, {a}.

Equating the kinetic energy expressed in each frame, we find an expression for the 6-by-6

spatial inertia matrix expressed in the {a} frame, G_a, in terms of the G_b matrix and

the transform T_ba expressing the {a} frame in the {b} frame.

With this inertia matrix, we find that the equation of motion in terms of the wrench

F_a, the spatial inertia matrix G_a, and the twist V_a has the same form as it does in

the {b} frame.

We usually prefer to write the equations in terms of a center of mass frame {b}, however.

To wrap up, we have derived the inverse dynamics for a rigid body: given the twist and acceleration,

we can calculate the wrench needed to generate this motion.

We can also write fsthe forward dynamics, which takes the current twist and the applied

wrench and calculates the acceleration.

In the next video we will use these results to derive the recursive Newton-Euler algorithm

for the inverse dynamics of a robot.