Download as pdf or txt
Download as pdf or txt
You are on page 1of 28

7th Lecture

Advanced Robotic Kinematics and Dynamics (ARKaD)


LAGRANGE Dynamics
Contents of the 7th Lecture

1 Dynamic Models

2 LAGRANGE Formulation

3 Properties of the Dynamic Model

Slide 2 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Dynamic Models

TCP: Tool Center Point


Deriving a mathematical description of the dynamic behavior of a manipulator
𝛿5 Path is important in many fields of robotics:
𝛿4 • Simulation
𝛿3 • Structural design
𝒓𝑃, 𝑛𝒐 • Control and motion planning

𝛿6 The aim of a dynamic model is to obtain a (computer processable) system of equations


characterizing the manipulator’s dynamic behaviour.
There are two common methods to derive a dynamic model:
• LAGRANGE formulation: energy formulation
𝛿2
𝛿1 • NEWTON-EULER formulation: dynamic equilibrium formulation
The dynamic model maps the joint actuator torques to the motion of the manipulator:

𝑴 𝒒 𝒒ሷ + 𝑮 𝒒, 𝒒ሶ 𝒒ሶ + 𝒈 𝒒 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ − 𝑱𝑇 𝒒 𝒉𝑒

Slide 3 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Contents of the 7th Lecture

1 Dynamic Models

2 LAGRANGE Formulation

3 Properties of the Dynamic Model

Slide 4 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Basic LAGRANGIAN and LAGRANGE Equations

The LAGRANGE formulation is a formulation for classical mechanic problems:


𝑇
• It is a function of the generalized joint coordinates 𝒒 = 𝑞1 𝑞2 𝑞3 𝑞4 ⋯ 𝑞𝑛
• It is based on the kinetic and potential energy of the system
• It is an alternative to the classical Newton-Euler formulation
The LAGRANGIAN function is defined as:
ℒ 𝒒, 𝒒ሶ = 𝐸𝑘𝑖𝑛 𝒒, 𝒒ሶ − 𝐸𝑝𝑜𝑡 𝒒

where 𝐸𝑘𝑖𝑛 denotes the total kinetic energy,


while 𝐸𝑝𝑜𝑡 denotes the total potential energy of the system.

The equations of motion can be derived independently of the reference frame using the LAGRANGE equations:
𝑑 𝜕ℒ 𝜕ℒ
− =𝝃
𝑑𝑡 𝜕𝒒ሶ 𝜕𝒒

𝝃 is the vector of non-conservatives generalized forces acting on the system mapped to the joint space:
• generalized joint actuator forces
• generalized joint friction forces The LAGRANGE equations establish the
relationship between the generalized forces and
• generalized end-effector process forces
the joint positions, velocities and accelerations.

Slide 5 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
0
𝝎𝑖
Kinetic Energy 0
𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂
𝑖
0
The kinetic energy is derived from the manipulator’s motion. 𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖
The total kinetic energy of a manipulator with 𝑛 rigid links is given by 𝐶𝐺
𝑑𝑉
the kinetic energy of each link 𝑖 𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 and of each motor 𝑖 𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖 : 𝑃
𝑛
𝐸𝑘𝑖𝑛 = ෍ 𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 + 𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖
𝑖=1 0 0
𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝒓𝑃, 0𝑂
The kinetic energy of link 𝑖 is given by its inertia, integrated across the whole volume of the link: 𝑖
1 0 ሶ𝑇
𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 = න 𝒓𝑃, 0𝑂 0𝒓ሶ 𝑃, 0𝑂 𝜌𝑑𝑉 𝒆𝑧, 0𝑂
2 𝑉𝑙𝑖𝑛𝑘
𝑖
𝒆𝑦, 0𝑂
𝒓ሶ 𝑃, 0𝑂 is the translational velocity of point 𝑃 on link 𝑖, and 𝜌 the density of an elementary volume particle. 0𝑂
0
𝒆𝑥, 0𝑂
The position vector to the center of gravity 𝐶𝐺 of link 𝑖 is: Kinematic representation of a link
0
1 0
𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 = න 𝒓 0 𝜌𝑑𝑉
𝑖 𝑚𝑙𝑖𝑛𝑘𝑖 𝑉𝑙𝑖𝑛𝑘 𝑃, 𝑂
𝑖

The location of an infinitesimal particle expressed in the base frame is:


0
𝒓𝑃, 0𝑂 = 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 0𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖 𝑖
Now the translational velocity of point 𝑷 can be expressed using the angular velocity 0𝝎𝑖 :
0
𝒓ሶ 𝑃, 0𝑂 = 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 0𝝎𝑖 × 0𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 = 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 𝑺 0𝝎𝑖 0𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖 𝑖 𝑖 𝑖

Slide 6 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
0
𝝎𝑖
Kinetic Energy Contributions 0
𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂
𝑖
From previous slide: 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖
1 0 ሶ𝑇 𝐶𝐺
𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 = න 𝒓 0 0𝒓ሶ 0 𝜌𝑑𝑉 𝑑𝑉
2 𝑉𝑙𝑖𝑛𝑘 𝑃, 𝑂 𝑃, 𝑂
𝑖 𝑃
0 0 0 0
𝒓ሶ 𝑃, 0𝑂 = 𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 𝑺 𝝎𝑖 𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖 𝑖

0 ሶ𝑇
𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 2 ∙ 0𝒓ሶ 𝑇𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝑺 0𝝎𝑖 0𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0 0
1 𝑖 𝑖
𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝒓𝑃, 0𝑂
𝑖 𝑖
This leads to: 𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 = න 𝜌𝑑𝑉 𝑖
2 𝑉𝑙𝑖𝑛𝑘 + 0𝒓𝑇𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝑺𝑇 0𝝎𝑖 𝑺 0𝝎𝑖 0𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖 𝑖 𝑖
𝒆𝑧, 0𝑂
0
Using 𝒓ሶ 𝑃, 0𝑂 , the equation of the kinetic energy for link 𝑖 offers three (two) contributions:
0 𝒆𝑦, 0𝑂
1 0 ሶ𝑇
1 𝑂
Translational: 𝐸𝑘𝑖𝑛,𝑡𝑟𝑎𝑛𝑠,𝑙𝑖𝑛𝑘𝑖 = න 𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝜌𝑑𝑉 = 𝑚𝑙𝑖𝑛𝑘𝑖 0𝒓ሶ 𝑇𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝒆𝑥, 0𝑂
2 𝑉𝑙𝑖𝑛𝑘 𝑖 𝑖 2 𝑖 𝑖 Kinematic representation of a link
𝑖

1 1
Mutual: 𝐸𝑘𝑖𝑛,𝑚𝑢𝑡,𝑙𝑖𝑛𝑘𝑖 = ∙ 2 ∙ 0𝒓ሶ 𝑇𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝑺 0𝝎𝑖 න 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉 = ∙ 2 ∙ 0𝒓ሶ 𝑇𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝑺 0𝝎𝑖 න 0
𝒓𝑃, 0𝑂 − 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝜌𝑑𝑉 = 0
2 𝑖 𝑉𝑙𝑖𝑛𝑘
𝑖 2 𝑖 𝑉𝑙𝑖𝑛𝑘 𝑖
𝑖 𝑖

1 1
Rotational: 𝐸𝑘𝑖𝑛,𝑟𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 = න 𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝑺𝑇 0𝝎𝑖 𝑺 0𝝎𝑖 0𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉 = 0𝝎𝑇𝑖 න
0 𝑇
𝑺𝑇 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝑺 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉 0
𝛚i
2 𝑉𝑙𝑖𝑛𝑘 𝑖 𝑖 2 𝑉𝑙𝑖𝑛𝑘
𝑖 𝑖
𝑖 𝑖

0 − 0𝑟𝑧;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0
𝑟𝑦;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖 𝑖
0 0 0 0 0 0 0
where the property 𝑺 𝝎𝑖 𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 = −𝑺 𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝝎𝑖 with the skew-matrix 𝑺 𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 = 𝑟𝑧;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0 − 𝑟𝑥;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 was used.
𝑖 𝑖 𝑖 𝑖 𝑖
− 0𝑟𝑦;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0
𝑟𝑥;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0
𝑖 𝑖

Slide 7 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Rotational Energy Contribution
10 𝑇
𝐸𝑘𝑖𝑛,𝑟𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 = 𝝎𝑖 න 𝑺𝑇 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝑺 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉 0
𝝎𝑖
2 𝑉𝑙𝑖𝑛𝑘
𝑖 𝑖
𝑖

Defining the integral as the inertia tensor relative to the center of mass of link 𝑖 expressed in the base frame as
0
𝑰𝑙𝑖𝑛𝑘𝑖 = න 𝑺𝑇 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝑺 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉
𝑖 𝑖
𝑉𝑙𝑖𝑛𝑘𝑖

0 2 2
න 𝑟𝑦;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 + 0𝑟𝑧;𝑃,𝐶𝐺 𝑙𝑖𝑛𝑘
𝜌𝑑𝑉 − න 0𝑟𝑥;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0𝑟𝑦;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉 − න 0𝑟𝑥;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0𝑟𝑧;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉
𝑖 𝑖 𝑖 𝑖 𝑖 𝑖

0 2 2
= ∗ න 𝑟𝑥;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 + 0𝑟𝑧;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝜌𝑑𝑉 − න 0𝑟𝑦;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 0𝑟𝑧;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 𝜌𝑑𝑉
𝑖 𝑖 𝑖 𝑖

0 2 2
∗ ∗ න 𝑟𝑥;𝑃,𝐶𝐺𝑙𝑖𝑛𝑘 + 0𝑟𝑦;𝑃,𝐶𝐺 𝑙𝑖𝑛𝑘
𝜌𝑑𝑉
𝑖 𝑖

Symmetric elements

Now the rotational energy contribution can alternatively be expressed as:


1
𝐸𝑘𝑖𝑛,𝑟𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 = 0𝝎𝑇𝑖 0𝑰𝑙𝑖𝑛𝑘𝑖 0𝝎𝑖
2

Slide 8 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
0
𝝎𝑖
The Constant Inertia Tensor 0
𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂
𝑖
0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑖
𝐶𝐺
𝑑𝑉
𝑃

The rotational energy contribution of link i can be expressed through the inertia tensor as:
0 0
1 𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝒓𝑃, 0𝑂
𝐸𝑘𝑖𝑛,𝑟𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 = 0𝝎𝑇𝑖 0𝑰𝑙𝑖𝑛𝑘𝑖 0𝝎𝑖 𝑖
2
𝒆𝑧, 0𝑂
Given the rotation matrix 0𝑹𝑖 from link i to the base frame,
the inertia tensor can be expressed as: 0 𝒆𝑦, 0𝑂
𝑂
𝑰𝑙𝑖𝑛𝑘𝑖 = 0𝑹𝑖 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖
0 𝒆𝑥, 0𝑂
Kinematic representation of a link
Other than the inertia tensor referenced to the base frame,
the inertia tensor referenced to the link frame i, 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 , is constant.

Summing up the translational and rotational contributions while using above rotation matrices yields:
1 1
𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 = 𝐸𝑘𝑖𝑛,𝑡𝑟𝑎𝑛𝑠,𝑙𝑖𝑛𝑘𝑖 + 𝐸𝑘𝑖𝑛,𝑟𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 = 𝑚𝑙𝑖𝑛𝑘𝑖 0𝒓ሶ 𝑇𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 0𝝎𝑇𝑖 0𝑹𝑖 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖 0𝝎𝑖
2 𝑖 𝑖 2

Slide 9 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Total Kinetic Energy of a Link LAGRANGE equations:
In order to allow for the differentiation of the LAGRANGIAN, the total kinetic energy equation 𝑑 𝜕ℒ 𝜕ℒ
− =𝝃
has to be expressed in generalized joint coordinates: 𝑑𝑡 𝜕 𝒒ሶ 𝜕𝒒
0
𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 = 𝑱𝐺𝑃𝑙𝑖𝑛𝑘 𝒒 𝒒ሶ
𝑖 𝑖
0 ൡ = 𝑱𝐺𝑙𝑖𝑛𝑘 𝒒 𝒒ሶ
𝝎𝑖 = 𝑱𝐺𝑂𝑙𝑖𝑛𝑘 𝒒 𝒒ሶ 𝑖
From Fifth Lecture
𝑖

The geometric Jacobian of link 𝑖 is as follows Prismatic Joint


𝑱𝐺𝑃𝑙𝑖𝑛𝑘 𝒋𝐺𝑃1𝑙𝑖𝑛𝑘 ⋯ 𝒋𝐺𝑃𝑖𝑙𝑖𝑛𝑘 𝟎 ⋯ 𝟎 𝒋𝐺𝑃𝑗 0
𝒆𝑧, 𝑗−1𝑂
𝑱𝐺𝑙𝑖𝑛𝑘 = 𝑖
= 𝑖 𝑖 =
𝑖 𝑱𝐺𝑂𝑙𝑖𝑛𝑘 𝒋𝐺𝑂1𝑙𝑖𝑛𝑘 ⋯ 𝒋𝐺𝑂𝑖𝑙𝑖𝑛𝑘 𝟎 ⋯ 𝟎 𝒋𝐺𝑂𝑗 𝟎
𝑖 𝑖 𝑖

Given the unit vector of axis 𝑧 in frame 𝑗 − 1 0𝒆𝑧,𝑗−1𝑂 the Jacobian columns can be computed by: Revolute Joint
0 0
0
𝒆𝑧, 𝑗−1𝑂 Prismatic Joint 𝒋𝐺𝑃𝑗 𝒆𝑧, 𝑗−1𝑂 × 𝒓𝐸, 0𝑂 − 0𝒓 𝑗−1𝑂, 0𝑂
=
𝒋𝐺𝑃𝑗𝑙𝑖𝑛𝑘 = ቐ 0 𝒋𝐺𝑂𝑗 0
𝒆𝑧, 𝑖−1𝑂
0 0
𝑖 𝒆𝑧, 𝑗−1𝑂 × 𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 − 𝒓 𝑗−1𝑂, 0𝑂 Revolute Joint
𝑖

𝟎 Prismatic Joint
𝒋𝐺𝑂𝑗𝑙𝑖𝑛𝑘 = ൝ 0𝒆 Revolute Joint
𝑖 𝑧, 𝑗−1𝑂
Thus the total kinetic energy of link 𝑖 can be written as:
1 1
𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 = 𝑚𝑙𝑖𝑛𝑘𝑖 0𝒓ሶ 𝑇𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 0𝝎𝑇𝑖 0𝑹𝑖 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖 0𝝎𝑖
2 𝑖 𝑖 2
1 1
= 𝑚𝑙𝑖𝑛𝑘𝑖 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑃𝑙𝑖𝑛𝑘 𝑱𝐺𝑃𝑙𝑖𝑛𝑘 𝒒ሶ + 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑂𝑙𝑖𝑛𝑘 0𝑹𝑖 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖 𝑱𝐺𝑂𝑙𝑖𝑛𝑘 𝒒ሶ
2 𝑖 𝑖 2 𝑖 𝑖

Slide 10 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Total Kinetic Energy Contribution of a Motor I
The kinetic energy contribution of the motor of joint 𝑖 on link 𝑖 − 1 can be expressed similar to the contribution of a link:
1 1
𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖 = 𝑚𝑚𝑜𝑡𝑖 0𝒓ሶ 𝑇𝐶𝐺𝑚𝑜𝑡 , 0𝑂 0𝒓ሶ 𝐶𝐺𝑚𝑜𝑡 , 0𝑂 + 0𝝎𝑇𝑚𝑜𝑡𝑖 0𝑰𝑚𝑜𝑡𝑖 0𝝎𝑚𝑜𝑡𝑖
2 𝑖 𝑖 2
0
𝝎𝑚𝑜𝑡𝑖
𝑚𝑚𝑜𝑡𝑖 is the mass of the motor 𝑖
motor 𝑖
0
0 𝒓ሶ 𝐶𝐺𝑚𝑜𝑡 , 0𝑂
• 𝒓ሶ 𝐶𝐺𝑚𝑜𝑡 , 0𝑂 is the translational velocity of the center of mass of the motor 𝑖
𝑖

0
• 𝑰𝑚𝑜𝑡𝑖 is the inertia tensor of the motor relative to its center of mass
0
• 𝝎𝑚𝑜𝑡𝑖 is the angular velocity of the motor

A motor is generally connected to the joint 𝑖 by a rigid transmission. link 𝑖 − 1


Thus, the angular velocity of the motor 𝑖 can be obtained from the gear ratio 𝑖𝑚𝑜𝑡𝑖 : joint 𝑖
0
𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂

𝛿𝑚𝑜𝑡 = 𝑖𝑚𝑜𝑡𝑖 𝑞ሶ 𝑖 𝑖
𝑖
𝒆𝑧, 0𝑂
Thus the total angular velocity of the motor 𝑖 can be expressed by:
0
𝝎𝑚𝑜𝑡𝑖 = 0𝝎𝑖−1 + 𝑖𝑚𝑜𝑡𝑖 𝑞ሶ 𝑖 0𝒆𝑧,𝑚𝑜𝑡𝑖 0 𝒆𝑦, 0𝑂
𝑂
𝒆𝑥, 0𝑂
Here 0𝝎𝑖−1 is the angular velocity of link 𝑖 − 1 , on which the motor 𝑖 is located,
Kinematic representation of an actuator
and 0𝒆𝑧,𝑚𝑜𝑡𝑖 is the unit vector along the motor axis.

Slide 11 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Total Kinetic Energy of a Motor (Actuator) II
Similar to the link‘s kinetic energy contribution, the kinetic energy of the motor 𝑖 can be expressed as a function of the joint variables:
0
𝒓ሶ 𝐶𝐺𝑚𝑜𝑡 , 0𝑂 = 𝑱𝐺𝑃𝑚𝑜𝑡 𝒒ሶ
𝑖 𝑖
0 ൡ = 𝑱𝐺𝑚𝑜𝑡 𝒒 𝒒ሶ
𝝎𝑚𝑜𝑡𝑖 = 𝑱𝐺𝑂𝑚𝑜𝑡 𝒒ሶ 𝑖
𝑖

The Jacobian is similar to the one for the link: From Fifth Lecture
𝑱𝐺𝑃𝑚𝑜𝑡 𝒋𝐺𝑃1𝑚𝑜𝑡 ⋯ 𝒋𝐺𝑃𝑖−1𝑚𝑜𝑡 Prismatic Joint
𝑖 𝑖 𝑖
𝟎 𝟎 ⋯ 𝟎
𝑱𝐺𝑚𝑜𝑡 = = 𝒋𝐺𝑃𝑖 0
𝒆𝑧, 𝑖−1𝑂
𝑖 𝑱𝐺𝑂𝑚𝑜𝑡 𝒋𝐺𝑂1𝑚𝑜𝑡 ⋯ 𝒋𝐺𝑂𝑖−1𝑚𝑜𝑡 𝒋𝐺𝑂𝑖𝑚𝑜𝑡𝑖 𝟎 ⋯ 𝟎 =
𝑖 𝑖 𝑖 𝒋𝐺𝑂𝑖 𝟎
with 𝒋𝐺𝑃𝑖𝑚𝑜𝑡 = 𝟎, because the motor’s center of mass is considered to be on its axis of rotation. Revolute Joint
𝑖

Given the unit vector 0𝒆𝑧,𝑗−1𝑂 of axis 𝑧 in frame 𝑗 − 1 the Jacobian columns can be computed by: 𝒋𝐺𝑃𝑖
0
𝒆𝑧, 𝑖−1𝑂 × 0
𝒓𝐸, 0𝑂 − 0𝒓 𝑖−1𝑂, 0𝑂
=
0 𝒋𝐺𝑂𝑖 0
𝒆𝑧, 𝑖−1𝑂
𝒆𝑧, 𝑗−1𝑂 Prismatic Joint
𝒋𝐺𝑃𝑗𝑚𝑜𝑡 = ቐ 0 0
𝑖 𝒆𝑧, 𝑗−1𝑂 × 𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂 − 0𝒓 𝑗−1𝑂, 0𝑂 Revolute Joint
𝑖

𝒋𝐺𝑂𝑗𝑙𝑖𝑛𝑘 for all links 𝑗 = 1, … , 𝑖 − 1


𝑖
𝒋𝐺𝑂𝑗𝑚𝑜𝑡 = ቐ
𝑖 𝑖𝑚𝑜𝑡𝑖 0𝒆𝑧,𝑚𝑜𝑡𝑖 for link 𝑗 = 𝑖

Thus the total kinetic energy of motor 𝑖 is:


1 1
𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖 = 𝑚𝑚𝑜𝑡𝑖 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑃𝑚𝑜𝑡 𝑱𝐺𝑃𝑚𝑜𝑡 𝒒ሶ + 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑂𝑚𝑜𝑡 0𝑹𝑚𝑜𝑡𝑖 𝑚𝑜𝑡𝑖 𝑰𝑚𝑜𝑡𝑖 0𝑹𝑇𝑚𝑜𝑡𝑖 𝑱𝐺𝑂𝑚𝑜𝑡 𝒒ሶ
2 𝑖 𝑖 2 𝑖 𝑖

Slide 12 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Total Kinetic Energy

Two kinetic energy contributions for each link:


1 1
𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 = 𝑚𝑙𝑖𝑛𝑘𝑖 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑃𝑙𝑖𝑛𝑘 𝑱𝐺𝑃𝑙𝑖𝑛𝑘 𝒒ሶ + 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑂𝑙𝑖𝑛𝑘 0𝑹𝑖 𝑖 𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖 𝑱𝐺𝑂𝑙𝑖𝑛𝑘 𝒒ሶ
2 𝑖 𝑖 2 𝑖 𝑖

1 1
𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖 = 𝑚𝑚𝑜𝑡𝑖 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑃𝑚𝑜𝑡 𝑱𝐺𝑃𝑚𝑜𝑡 𝒒ሶ + 𝒒ሶ 𝑇 𝑱𝑇𝐺𝑂𝑚𝑜𝑡 0𝑹𝑚𝑜𝑡𝑖 𝑚𝑜𝑡𝑖 𝑰𝑚𝑜𝑡𝑖 0𝑹𝑇𝑚𝑜𝑡𝑖 𝑱𝐺𝑂𝑚𝑜𝑡 𝒒ሶ
2 𝑖 𝑖 2 𝑖 𝑖

The total kinetic energy of the manipulator with its 𝑛 links and motors can be expressed as:
𝑛 1
𝐸𝑘𝑖𝑛 = ෍ 𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 + 𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖 = 𝒒ሶ 𝑇 𝑴 𝒒 𝒒ሶ
𝑖=1 2
with the inertia matrix:
𝑛
𝑴 𝒒 =෍ 𝑚𝑙𝑖𝑛𝑘𝑖 𝑱𝑇𝐺𝑃𝑙𝑖𝑛𝑘 𝑱𝐺𝑃𝑙𝑖𝑛𝑘 + 𝑱𝑇𝐺𝑂𝑙𝑖𝑛𝑘 0𝑹𝑖 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖 𝑱𝐺𝑂𝑙𝑖𝑛𝑘 + 𝑚𝑚𝑜𝑡𝑖 𝑱𝑇𝐺𝑃𝑚𝑜𝑡 𝑱𝐺𝑃𝑚𝑜𝑡 + 𝑱𝑇𝐺𝑂𝑚𝑜𝑡 0𝑹𝑚𝑜𝑡𝑖 𝑚𝑜𝑡𝑖 𝑰𝑚𝑜𝑡𝑖 0𝑹𝑇𝑚𝑜𝑡𝑖 𝑱𝐺𝑂𝑚𝑜𝑡
𝑖=1 𝑖 𝑖 𝑖 𝑖 𝑖 𝑖 𝑖 𝑖

Properties of the inertia matrix:


• Symmetric
• Positive definite
• Configuration-dependent but independent of the joint velocities

Slide 13 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Total Potential Energy

0
𝝎𝑖
0
Similar to the kinetic energy, the total potential energy is given 𝒓ሶ 𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂
𝑖
by the sum of terms related to the 𝑛 links and motors of the manipulator: 0
𝒓𝑃,𝐶𝐺𝑙𝑖𝑛𝑘
𝑛 𝑖
𝐸𝑝𝑜𝑡 = ෍ 𝐸𝑝𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 + 𝐸𝑝𝑜𝑡,𝑚𝑜𝑡𝑖 𝐶𝐺
𝑑𝑉
𝑖=1
𝑃
Assuming rigid links, the total link contribution is:

𝐸𝑝𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 = − 0 𝒈𝑇0 න 0
𝒓𝑃, 0𝑂 𝜌𝑑𝑉 = −𝑚𝑙𝑖𝑛𝑘𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0
𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 0
𝒓𝑃, 0𝑂
𝑖
𝑉𝑙𝑖𝑛𝑘𝑖 𝑖

Here 0 𝒈0 is the gravitational acceleration vector with respect to frame 0. 𝒆𝑧, 0𝑂


0 𝑇 0
• In general 𝒈0 = 0 0 −𝑔 , when 𝒆𝑧,0 is a vertical axis.
0 𝒆𝑦, 0𝑂
The total motor contribution is: 𝑂
𝒆𝑥, 0𝑂
Kinematic representation of a link
𝐸𝑝𝑜𝑡,𝑚𝑜𝑡𝑖 = −𝑚𝑚𝑜𝑡𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂
𝑖

The total potential energy is given by:


𝑛
𝐸𝑝𝑜𝑡 = − ෍ 𝑚𝑙𝑖𝑛𝑘𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 𝑚𝑚𝑜𝑡𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂
𝑖=1 𝑖 𝑖

Slide 14 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Equations of Motion I

The LAGRANGIAN for a manipulator is


1 𝑇
ℒ 𝒒, 𝒒ሶ = 𝐸𝑘𝑖𝑛 𝒒, 𝒒ሶ − 𝐸𝑝𝑜𝑡 𝒒 =𝒒ሶ 𝑴 𝒒 𝒒ሶ − 𝐸𝑝𝑜𝑡 𝒒
2
where the potential energy is independent of the joint velocities.
The equations of motions can be determined using the LAGRANGE equations:
𝑑 𝜕ℒ 𝜕ℒ
− =𝝃
𝑑𝑡 𝜕 𝒒ሶ 𝜕𝒒
𝝃 is the vector of non-conservatives generalized forces acting on the system mapped to the joint space.
The partial derivative of the LAGRANGIAN with respect to 𝒒ሶ yields:
1 𝑇
𝜕ℒ 𝜕 2 𝒒ሶ 𝑴 𝒒 𝒒ሶ
= = 𝑴 𝒒 𝒒ሶ
𝜕𝒒ሶ 𝜕 𝒒ሶ
Now the equations of motion can be formulated as:
𝑇 𝑇
1 𝜕 𝜕𝐸𝑝𝑜𝑡 𝒒
𝑴 𝒒 𝒒ሷ + 𝑴ሶ 𝒒 𝒒ሶ − 𝒒ሶ 𝑇 𝑴 𝒒 𝒒ሶ + =𝝃
2 𝜕𝒒 𝜕𝒒

Slide 15 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Analysing The Equations of Motion
The equations of motions are given by:
𝑇 𝑇 𝑛
1 𝜕 𝜕𝐸𝑝𝑜𝑡 𝒒
𝑴 𝒒 𝒒ሷ + 𝑴ሶ 𝒒 𝒒ሶ − 𝒒ሶ 𝑇 𝑴 𝒒 𝒒ሶ + = 𝝃 with 𝐸𝑝𝑜𝑡 = − ෍ 𝑚𝑙𝑖𝑛𝑘𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 𝑚𝑚𝑜𝑡𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂
2 𝜕𝒒 𝜕𝒒 𝑖=1 𝑖 𝑖

First two addends on the left side for a single link 𝑖:


𝑛 𝑛
𝑑𝑚𝑖𝑗 𝒒 𝑛 𝑛 𝑛 𝜕𝑚𝑖𝑗 𝒒
𝑴 𝒒 𝒒ሷ + 𝑴ሶ 𝒒 𝒒ሶ = ෍ 𝑚𝑖𝑗 𝒒 𝑞ሷ 𝑗 + ෍ 𝑞ሶ 𝑗 = ෍ 𝑚𝑖𝑗 𝒒 𝑞ሷ 𝑗 + ෍ ෍ 𝑞ሶ 𝑘 𝑞ሶ 𝑗
𝑖 𝑗=1 𝑗=1 𝑑𝑡 𝑗=1 𝑗=1 𝑘=1 𝜕𝑞𝑘
Third addend on the left side for a single link 𝑖:
𝑇
1 𝜕 1 𝑛 𝑛 𝜕𝑚𝑗𝑘 𝒒 𝑛 𝑛 1 𝜕𝑚𝑗𝑘 𝒒
− 𝒒ሶ 𝑇 𝑴 𝒒 𝒒ሶ =− ෍ ෍ 𝑞ሶ 𝑘 𝑞ሶ 𝑗 = ෍ ෍ − 𝑞ሶ 𝑘 𝑞ሶ 𝑗
2 𝜕𝒒 2 𝑗=1 𝑘=1 𝜕𝑞𝑖 𝑗=1 𝑘=1 2 𝜕𝑞𝑖
𝑖

Fourth addend on the left side for a single link 𝑖:

𝜕𝐸𝑝𝑜𝑡 𝑛 𝜕 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝜕 0𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂


𝑗 𝑗
= −෍ 𝑚𝑙𝑖𝑛𝑘𝑗 0 𝒈𝑇0 + 𝑚𝑚𝑜𝑡𝑗 0 𝒈𝑇0 = 𝑔𝑖 (𝒒)
𝜕𝑞𝑖 𝑗=1 𝜕𝑞𝑖 𝜕𝑞𝑖
Thus the equations of motions are given by:
𝑛 𝑛 𝑛 𝜕𝑚𝑖𝑗 1 𝜕𝑚𝑗𝑘
෍ 𝑚𝑖𝑗 𝒒 𝑞ሷ 𝑗 + ෍ ෍ ℎ𝑖𝑗𝑘 𝒒 𝑞ሶ 𝑘 𝑞ሶ 𝑗 + 𝑔𝑖 𝒒 = 𝜉𝑖 𝑖 = 1, … , 𝑛 with ℎ𝑖𝑗𝑘 = −
𝑗=1 𝑗=1 𝑘=1 𝜕𝑞𝑘 2 𝜕𝑞𝑖

Slide 16 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Physical Properties
𝑴 𝒒 𝒒ሷ + 𝑮 𝒒, 𝒒ሶ 𝒒ሶ + 𝒈 𝒒 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ − 𝑱𝑇 𝒒 𝒉𝑒

The equations of motions:


𝑛 𝑛 𝑛 𝜕𝑚𝑖𝑗 1 𝜕𝑚𝑗𝑘
෍ 𝑚𝑖𝑗 𝒒 𝑞ሷ 𝑗 + ෍ ෍ ℎ𝑖𝑗𝑘 𝒒 𝑞ሶ 𝑘 𝑞ሶ 𝑗 + 𝑔𝑖 𝒒 = 𝜉𝑖 𝑖 = 1, … , 𝑛 with ℎ𝑖𝑗𝑘 = −
𝑗=1 𝑗=1 𝑘=1 𝜕𝑞𝑘 2 𝜕𝑞𝑖
This representation has important physical properties regarding the left side of the equation:
• The coefficients 𝑚𝑖𝑖 denote the generalized inertia at the axis of joint 𝑖 assuming other joints are blocked translational
acceleration terms:
• The coefficients 𝑚𝑖𝑗 denote the effect of accelerations of joint 𝑖 on joint 𝑗 𝑴 𝒒 𝒒ሷ
• The terms ℎ𝑖𝑗𝑗 𝑞ሶ 𝑗2 denote the centrifugal effect induced on joint 𝑖 by joint 𝑗 Quadratic velocity
terms: 𝑮 𝒒, 𝒒ሶ 𝒒ሶ
• The terms ℎ𝑖𝑗𝑘 𝑞ሶ 𝑗 𝑞ሶ 𝑘 denote the Coriolis effect induced on joint 𝑖 by the velocities in joint 𝑗 and joint 𝑘

• The terms 𝑔𝑖 denote the moment at joint 𝑖 induced by gravity Only Configuration-
dependent terms: 𝒈 𝒒

Slide 17 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Joint Space Dynamic Model
𝑴 𝒒 𝒒ሷ + 𝑮 𝒒, 𝒒ሶ 𝒒ሶ + 𝒈 𝒒 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ − 𝑱𝑇 𝒒 𝒉𝑒

The non-conservative forces at the manipulator joints are given by 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ :


• 𝝉 is the vector of generalized joint forces
• 𝑪𝑣 𝒒ሶ is the viscous friction in the joints with the diagonal matrix 𝑪𝑣 containing the corresponding viscous friction coefficients
• 𝑭𝑑 sgn 𝒒ሶ is the COULOMB friction in the joints with the diagonal matrix 𝑭𝑑 containing the corresponding dry friction forces/torques
If external forces from the environment are applied to the end effector a portion of the joint torques 𝝉 is used to balance them:
• 𝜸𝑒 are the end-effector generalized forces in the operational space induced by the torques 𝝉
• 𝒉𝑒 are the end-effector generalized forces in the operational space induced from environmental influence
𝜸𝑒
𝑇 𝑇
In the joint space the following relation holds true: 𝑱 𝒒 𝜸𝑒 − 𝒉𝑒 = 𝝉 − 𝑱 𝒒 𝒉𝑒
The components of the matrix of gyroscope forces 𝑮 are given by (slide 16):
𝑛 𝑛 𝑛
𝒉𝑒
𝜕𝑚𝑖𝑗 1 𝜕𝑚𝑗𝑘
෍ 𝑔𝑖𝑗 𝑞ሶ 𝑗 = ෍ ෍ ℎ𝑖𝑗𝑘 𝑞ሶ 𝑘 𝑞ሶ 𝑗 , 𝑖 = 1, … , 𝑛 with ℎ𝑖𝑗𝑘 = −
𝑗=1 𝑗=1 𝑘=1 𝜕𝑞𝑘 2 𝜕𝑞𝑖
The components of the vector of generalized potential forces 𝒈 𝒒 are given by:

𝜕𝐸𝑝𝑜𝑡 𝑛 𝜕 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 𝜕 0𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂


𝑗 𝑗
= −෍ 𝑚𝑙𝑖𝑛𝑘𝑗 0 𝒈𝑇0 + 𝑚𝑚𝑜𝑡𝑗 0 𝒈𝑇0 = 𝑔𝑖 𝒒 , 𝑖 = 1, … , 𝑛
𝜕𝑞𝑖 𝑗=1 𝜕𝑞𝑖 𝜕𝑞𝑖

Slide 18 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Dry and Viscous Friction Matrices
𝑴 𝒒 𝒒ሷ + 𝑮 𝒒, 𝒒ሶ 𝒒ሶ + 𝒈 𝒒 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ − 𝑱𝑇 𝒒 𝒉𝑒

The non-conservative forces at the manipulator joints are given by 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ :


• 𝝉 is the vector of generalized joint forces
• 𝑪𝑣 𝒒ሶ is the viscous friction in the joints with the diagonal matrix 𝑪𝑣 containing the corresponding viscous friction coefficients
• 𝑭𝑑 sgn 𝒒ሶ is the COULOMB friction in the joints with the diagonal matrix 𝑭𝑑 containing the corresponding dry friction forces/torques
• 𝑪𝑣 has elements of two types on its diagonal
• 𝑐𝑣𝑖 translational viscous damping coefficient, if 𝑖 is a prismatic joint
• 𝑐𝜔𝑖 rotational viscous damping coefficient, if 𝑖 is a revolute joint 𝜸𝑒
• 𝑭𝑑 has elements of two types on its diagonal
• 𝐹𝑑𝑖 dry friction force, if 𝑖 is a prismatic joint
𝒉𝑒
• 𝑇𝑑𝑖 dry friction torque, if 𝑖 is a revolute joint

Slide 19 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Summary of LAGRANGE Formulation
1. The LAGRANGIAN function is defined as: ℒ 𝒒, 𝒒ሶ = 𝐸𝑘𝑖𝑛 𝒒, 𝒒ሶ − 𝐸𝑝𝑜𝑡 𝒒

2. The equations of motion can be derived from the LAGRANGIAN equations:


𝑑 𝜕ℒ 𝜕ℒ
− =𝝃
𝑑𝑡 𝜕𝒒ሶ 𝜕𝒒

3. The total kinetic energy of a manipulator with 𝑛 rigid links is given by:
𝑛 1 𝑇
𝐸𝑘𝑖𝑛 = ෍ 𝐸𝑘𝑖𝑛,𝑙𝑖𝑛𝑘𝑖 + 𝐸𝑘𝑖𝑛,𝑚𝑜𝑡𝑖 = 𝒒ሶ 𝑴 𝒒 𝒒ሶ
𝑖=1 2
with the mass matrix:
𝑛
𝑴 𝒒 =෍ 𝑚𝑙𝑖𝑛𝑘𝑖 𝑱𝑇𝐺𝑃𝑙𝑖𝑛𝑘 𝑱𝐺𝑃𝑙𝑖𝑛𝑘 + 𝑱𝑇𝐺𝑂𝑙𝑖𝑛𝑘 0𝑹𝑖 𝑖𝑰𝑙𝑖𝑛𝑘𝑖 0𝑹𝑇𝑖 𝑱𝐺𝑂𝑙𝑖𝑛𝑘 + 𝑚𝑚𝑜𝑡𝑖 𝑱𝑇𝐺𝑃𝑚𝑜𝑡 𝑱𝐺𝑃𝑚𝑜𝑡 + 𝑱𝑇𝐺𝑂𝑚𝑜𝑡 0𝑹𝑚𝑜𝑡𝑖 𝑚𝑜𝑡𝑖 𝑰𝑚𝑜𝑡𝑖 0𝑹𝑇𝑚𝑜𝑡𝑖 𝑱𝐺𝑂𝑚𝑜𝑡
𝑖=1 𝑖 𝑖 𝑖 𝑖 𝑖 𝑖 𝑖 𝑖

4. The total potential energy of a manipulator with 𝑛 rigid links is given by:
𝑛 𝑛
𝐸𝑝𝑜𝑡 = ෍ 𝐸𝑝𝑜𝑡,𝑙𝑖𝑛𝑘𝑖 + 𝐸𝑝𝑜𝑡,𝑚𝑜𝑡𝑖 = − ෍ 𝑚𝑙𝑖𝑛𝑘𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑙𝑖𝑛𝑘 , 0𝑂 + 𝑚𝑚𝑜𝑡𝑖 0 𝒈𝑇0 0𝒓𝐶𝐺𝑚𝑜𝑡 , 0𝑂
𝑖=1 𝑖=1 𝑖 𝑖

5. The equations of motion can be formulated as:


𝑇 𝑇
1 𝜕 𝜕𝐸𝑝𝑜𝑡 𝒒
𝑴 𝒒 𝒒ሷ + 𝑴ሶ 𝒒 𝒒ሶ − 𝒒ሶ 𝑇 𝑴 𝒒 𝒒ሶ + =𝝃
2 𝜕𝒒 𝜕𝒒
𝑴 𝒒 𝒒ሷ + 𝑮 𝒒, 𝒒ሶ 𝒒ሶ + 𝒈 𝒒 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ − 𝑱𝑇 𝒒 𝒉𝑒

Slide 20 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Contents of the 7th Lecture

1 Dynamic Models

2 LAGRANGE Formulation

3 Properties of the Dynamic Model

Slide 21 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Parameter Linearity in the Kinetic Energy Term
It can be shown without proof here that the kinetic energy is linear in the dynamic parameters:
1 1 1 2
𝐸𝑘𝑖𝑛 𝑖 = 𝑚𝑖 𝑖𝒓ሶ 𝑇𝑖𝑂, 0𝑂 𝑖𝒓ሶ 𝑖𝑂, 0𝑂 + 𝑖𝒓ሶ 𝑇𝑖𝑂, 0𝑂 𝑺 𝑖𝝎𝒊 𝑚𝑖 𝑖𝒓𝐶𝐺𝑖 , 𝑖𝑂 + 𝑖𝝎𝑇𝑖 𝑖𝑰෠𝑖 𝑖𝝎𝑖 +𝐼𝑚𝑜𝑡𝑖+1 𝑖𝑚𝑜𝑡𝑖+1 𝑞ሶ 𝑖+1 𝑖𝒆𝑇𝑧,𝑚𝑜𝑡𝑖+1 𝑖𝝎𝑖 + 𝑖𝑚𝑜𝑡 2
𝑞ሶ 𝑖+1 𝑖 = 1, … , 𝑛
2 2 2 𝑖+1

• Mass 𝑚𝑖 (one component)


• The first moment of inertia 𝑚𝑖 𝑖𝒓𝐶𝐺𝑖 , 𝑖𝑂 (three components)
• The inertia tensor 𝑖 𝑰෠𝑖 (six components)
• The motor inertia 𝐼𝑚𝑜𝑡𝑖+1 (one component)

Motor 1 is not yet considered. Its kinetic energy is also linear w.r.t. the motor inertia 𝐼𝑚𝑜𝑡1 :
1 2
𝐸𝑘𝑖𝑛 𝑚𝑜𝑡 = 𝑖𝑚𝑜𝑡 𝑞ሶ 2 𝐼
1 1 𝑚𝑜𝑡1
1 2
The potential energy term is also linear w.r.t. the mass 𝑚𝑖 and the first moment of inertia 𝑚𝑖 𝑖 𝒓𝐶𝐺𝑖 ,𝑖𝑂 :

𝐸𝑝𝑜𝑡 = −𝑚𝑖 0𝒈𝑇0 𝑖𝒓𝐶𝐺𝑖 , 0𝑂 = − 0𝒈𝑇0 𝑚𝑖 𝑖 𝒓 𝑖𝑂, 0𝑂 + 𝑚𝑖 𝑖 𝒓𝐶𝐺𝑖 ,𝑖𝑂


𝑖
For each link a constant 11 × 1 vector of the dynamic parameters can be arranged:
𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑇
𝝅𝑖 = 𝑚𝑖 𝑚𝑖 𝑖 𝑟𝑥;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑦;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖 𝑟𝑧;𝐶𝐺𝑖 , 𝑖𝑂 𝐼𝑥𝑥;𝑖 𝐼𝑥𝑦;𝑖 𝐼𝑥𝑧;𝑖 𝐼𝑦𝑦;𝑖 𝐼𝑦𝑧;𝑖 𝐼𝑧𝑧;𝑖 𝐼𝑚𝑜𝑡𝑖
Now the LAGRANGIAN of the system can be reformulated as:
𝑛 𝑛
ℒ 𝒒, 𝒒ሶ = ෍ 𝐸𝑘𝑖𝑛 𝑖 𝒒, 𝒒ሶ − 𝐸𝑝𝑜𝑡 𝑖 𝒒 =෍ 𝜷𝑇𝐸𝑘𝑖𝑛 𝑖 𝒒, 𝒒ሶ − 𝜷𝑇𝐸𝑝𝑜𝑡 𝒒 𝝅𝑖 𝑖 = 1, … , 𝑛
𝑖=1 𝑖=1 𝑖

Slide 22 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
LAGRANGIAN and Parameter Linearity
The Lagrangian of the system can be formulated as:
𝑛
ℒ 𝒒, 𝒒ሶ = ෍ 𝜷𝑇𝐸𝑘𝑖𝑛 𝒒, 𝒒ሶ − 𝜷𝑇𝐸𝑝𝑜𝑡 𝒒 𝝅𝑖 𝑖 = 1, … , 𝑛
𝑖 𝑖
𝑖=1

The equations of motion can be derived from the Lagrange equations:

𝜕 σ𝑛𝑖=1 𝜷𝑇𝐸𝑘𝑖𝑛 𝒒, 𝒒ሶ − 𝜷𝑇𝐸𝑝𝑜𝑡 𝒒 𝜕 σ𝑛𝑖=1 𝜷𝑇𝐸𝑘𝑖𝑛 𝒒, 𝒒ሶ − 𝜷𝑇𝐸𝑝𝑜𝑡 𝒒


𝑑 𝜕ℒ 𝜕ℒ 𝑑 𝑖 𝑖 𝑖 𝑖
− = 𝝅− 𝝅=𝝃
𝑑𝑡 𝜕 𝒒ሶ 𝜕𝒒 𝑑𝑡 𝜕 𝒒ሶ 𝜕𝒒
𝝅2 ⋯ 𝝅𝑛 𝑇
with the 𝝅 = 𝝅1
and 𝝃 the vector of non-conservatives generalized forces acting on the system mapped to the joint space.
Now the above equation can be written as a linear function of 𝝅 and 𝝃 with the 𝑛 × 11𝑛 regressor matrix 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ :
𝑇 𝑇 𝑇 𝝅1
𝜉1 𝒚11 𝒚12 ⋯ 𝒚1𝑛
𝜉
𝝃 = 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅 = 2 = 𝟎
𝑇
𝒚𝑇22 ⋯ 𝒚𝑇2𝑛 𝝅2
⋮ ⋮ ⋮ ⋱ ⋮ ⋮
𝜉𝑛 𝟎 𝑇
𝟎 𝑇
⋯ 𝒚𝑇𝑛𝑛 𝝅𝑛
𝑑 𝛿𝜷𝐸𝑘𝑖𝑛 𝑖 𝒒, 𝒒ሶ 𝛿𝜷𝐸𝑘𝑖𝑛 𝑖 𝒒, 𝒒ሶ 𝑑 𝛿𝜷𝐸𝑝𝑜𝑡 𝑖 𝒒 𝛿𝜷𝐸𝑝𝑜𝑡 𝒒
𝑖
with: 𝒚𝑖𝑗 = − − + and 𝒚𝑖𝑗 = 𝟎, if 𝑗 < 𝑖
𝑑𝑡 𝛿 𝑞ሶ 𝑖 𝛿𝑞𝑖 𝑑𝑡 𝛿 𝑞ሶ 𝑖 𝛿𝑞𝑖
𝟎
Slide 23 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves
Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Dynamic Parameter Identification
Linear function of 𝝅 and 𝝃 with the 𝑛 × 11𝑛 regressor matrix :
𝑇 𝑇 𝑇 𝝅1
𝜉1 𝒚11 𝒚12 ⋯ 𝒚1𝑛
𝜉 𝑇
𝒚𝑇22 ⋯ 𝒚𝑇2𝑛 𝝅2
𝝃 = 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅 = 2 = 𝟎 ⋮
⋮ ⋮ ⋮ ⋱ ⋮
𝜉𝑛 𝟎 𝑇
𝟎 𝑇 𝑇
⋯ 𝒚𝑛𝑛 𝝅𝑛

𝝃 is the vector of non-conservatives generalized forces acting on the system mapped to the joint space:
• generalized joint actuator forces (𝝉)
• generalized joint friction forces (−𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ )
• generalized end-effector process forces
Without generalized end-effector process forces:
𝑇 𝑇 𝑇 𝝅1
𝑐11 0 ⋯ 0 𝑓11 0 ⋯ 0 𝒚11 𝒚12 ⋯ 𝒚1𝑛
0 𝑐22 ⋯ 0 0 𝑓22 ⋯ 0 𝟎 𝑇
𝒚 𝑇
⋯ 𝒚𝑇2𝑛 𝝅2
𝝃 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ = 𝝉 − 𝒒ሶ − sgn 𝒒ሶ = 22
⋮ = 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅
⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮
0 0 ⋯ 𝑐𝑛𝑛 0 0 ⋯ 𝑓𝑛𝑛 𝟎 𝑇
𝟎 𝑇 𝑇
⋯ 𝒚𝑛𝑛 𝝅𝑛

This equation can be rewritten as 𝝉 = 𝝃 + 𝑪𝑣 𝒒ሶ + 𝑭𝑑 sgn 𝒒ሶ = 𝒀∗ 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅∗


with 𝝅∗ = 𝝅1∗ 𝝅2∗ ⋯ 𝝅∗𝑛
𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑇
and 𝝅∗𝑖 = 𝑚𝑖 𝑚𝑖 𝑖𝑟𝑥;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑦;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑧;𝐶𝐺𝑖 , 𝑖𝑂 𝐼𝑥𝑥;𝑖 𝐼𝑥𝑦;𝑖 𝐼𝑥𝑧;𝑖 𝐼𝑦𝑦;𝑖 𝐼𝑦𝑧;𝑖 𝐼𝑧𝑧;𝑖 𝐼𝑚𝑜𝑡𝑖 𝑐𝑖𝑖 𝑓𝑖𝑖
Slide 24 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves
Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Dynamic Parameter Identification
Without generalized end-effector process forces:
𝑇 𝑇 𝑇 𝝅1
𝑐11 0 ⋯ 0 𝑓11 0 ⋯ 0 𝒚11 𝒚12 ⋯ 𝒚1𝑛
0 𝑐22 ⋯ 0 0 𝑓22 ⋯ 0 𝑇
𝒚𝑇22 ⋯ 𝒚𝑇2𝑛 𝝅2
𝝃 = 𝝉 − 𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ = 𝝉 − 𝒒ሶ − sgn 𝒒ሶ = 𝟎 ⋮ = 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅
⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮
0 0 ⋯ 𝑐𝑛𝑛 0 0 ⋯ 𝑓𝑛𝑛 𝟎𝑇 𝟎𝑇 ⋯ 𝒚𝑇𝑛𝑛 𝝅𝑛

This equation can be rewritten as 𝝉 = 𝝃 + 𝑪𝑣 𝒒ሶ + 𝑭𝑑 sgn 𝒒ሶ = 𝒀∗ 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅∗


with 𝝅∗ = 𝝅1∗ 𝝅2∗ ⋯ 𝝅∗𝑛
𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑖መ 𝑇
and 𝝅∗𝑖 = 𝑚𝑖 𝑚𝑖 𝑖 𝑟𝑥;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑦;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑧;𝐶𝐺𝑖 , 𝑖𝑂 𝐼𝑥𝑥;𝑖 𝐼𝑥𝑦;𝑖 𝐼𝑥𝑧;𝑖 𝐼𝑦𝑦;𝑖 𝐼𝑦𝑧;𝑖 𝐼𝑧𝑧;𝑖 𝐼𝑚𝑜𝑡𝑖 𝑐𝑖𝑖 𝑓𝑖𝑖

Accordingly 𝒀∗ 𝒒, 𝒒,ሶ 𝒒ሷ can be computed from:


∗ 𝑇 ∗ 𝑇 ∗ 𝑇
𝒚11 𝒚12 ⋯ 𝒚1𝑛
𝑇 ∗ 𝑇 ∗ 𝑇
∗ ሶ ሷ
𝒀 𝒒, 𝒒, 𝒒 = 𝟎 𝒚22 ⋯ 𝒚2𝑛
⋮ ⋮ ⋱ ⋮
𝟎𝑇 𝟎𝑇 ⋯ 𝒚𝑛𝑛∗ 𝑇


with 𝒚𝑖𝑖 = 𝒚𝑇𝑖𝑖 𝑞ሶ 𝑖 sgn 𝑞ሶ 𝑇


and 𝒚𝑖𝑗 = 𝒚𝑇𝑖𝑖 0 0 𝑇
, if 𝑗 > 𝑖

Slide 25 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Summary of Properties of Dynamic Models
1. It can be shown without proof here that the kinetic energy is linear in the dynamic parameters
1 1 1 2
𝐸𝑘𝑖𝑛 𝑖 = 𝑚𝑖 𝑖𝒓ሶ 𝑇𝑖𝑂, 0𝑂 𝑖𝒓ሶ 𝑖𝑂, 0𝑂 + 𝑖𝒓ሶ 𝑇𝑖𝑂, 0𝑂 𝑺 𝑖𝝎𝒊 𝑚𝑖 𝑖𝒓𝐶𝐺𝑖 , 𝑖𝑂 + 𝑖𝝎𝑇𝑖 𝑖𝑰෠𝑖 𝑖𝝎𝑖 +𝐼𝑚𝑜𝑡𝑖+1 𝑖𝑚𝑜𝑡𝑖+1 𝑞ሶ 𝑖+1 𝑖𝒆𝑇𝑧,𝑚𝑜𝑡𝑖+1 𝑖𝝎𝑖 + 𝑖𝑚𝑜𝑡 2
𝑞ሶ 𝑖+1
2 2 2 𝑖+1

2. For each link a constant vector of the dynamic parameters can be arranged:
𝝅𝑖 = 𝑚𝑖 𝑚𝑖 𝑖𝑟𝑥;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑦;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑧;𝐶𝐺𝑖 , 𝑖𝑂 𝑖𝐼መ𝑥𝑥;𝑖 𝑖𝐼መ𝑥𝑦;𝑖 𝑖መ
𝐼𝑥𝑧;𝑖 𝑖መ
𝐼𝑦𝑦;𝑖 𝑖መ
𝐼𝑦𝑧;𝑖 𝑖መ
𝐼𝑧𝑧;𝑖 𝐼𝑚𝑜𝑡𝑖

3. Based on the LAGRANGE Equations the vector of non-conservatives generalized forces 𝝃 can be written as a linear function of 𝝅
with the 𝑛 × 11𝑛 regressor matrix 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ
𝑇 𝑇 𝑇 𝝅1
𝜉1 𝒚11 𝒚12 ⋯ 𝒚1𝑛
𝜉
𝝃 = 𝒀 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅 = 2 = 𝟎
𝑇
𝒚𝑇22 ⋯ 𝒚𝑇2𝑛 𝝅2
⋮ ⋮ ⋮ ⋱ ⋮ ⋮
𝜉𝑛 𝟎𝑇 𝟎𝑇 ⋯ 𝒚𝑇𝑛𝑛 𝝅𝑛
4. If joint friction forces (−𝑪𝑣 𝒒ሶ − 𝑭𝑑 sgn 𝒒ሶ ) are assumed above vector 𝝅𝑖 can be extended to:
𝑇
𝝅𝑖∗ = 𝑚𝑖 𝑚𝑖 𝑖𝑟𝑥;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑦;𝐶𝐺𝑖 , 𝑖𝑂 𝑚𝑖 𝑖𝑟𝑧;𝐶𝐺𝑖 , 𝑖𝑂 𝑖𝐼መ𝑥𝑥;𝑖 𝑖𝐼መ𝑥𝑦;𝑖 𝑖𝐼መ𝑥𝑧;𝑖 𝑖𝐼መ𝑦𝑦;𝑖 𝑖𝐼መ𝑦𝑧;𝑖 𝑖መ
𝐼𝑧𝑧;𝑖 𝐼𝑚𝑜𝑡𝑖 𝑐𝑖𝑖 𝑓𝑖𝑖

5. Now the generalized joint actuator forces 𝝉 can be expressed by:


𝝉 = 𝝃 + 𝑪𝑣 𝒒ሶ + 𝑭𝑑 sgn 𝒒ሶ = 𝒀∗ 𝒒, 𝒒,ሶ 𝒒ሷ 𝝅∗ : ∗ 𝑇 ∗ 𝑇 ∗ 𝑇
𝒚11 𝒚12 ⋯ 𝒚1𝑛
𝑇 ∗ 𝑇 ∗ 𝑇
6. The new 𝑛 × 13𝑛 regressor matrix is given by: 𝒀∗ 𝒒, 𝒒,ሶ 𝒒ሷ = 𝟎 𝒚22 ⋯ 𝒚2𝑛
⋮ ⋮ ⋱ ⋮
𝑇 𝑇 ∗ 𝑇
𝟎 𝟎 ⋯ 𝒚𝑛𝑛
Slide 26 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves
Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Contents of the 7th Lecture

1 Dynamic Models

2 LAGRANGE Formulation

3 Properties of the Dynamic Model

Slide 27 / 28 Lagrange | Univ.-Prof. Dr.-Ing. Dr. h. c. B. Corves


Institute for Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Thank you
for your kind attention

Contact:
Institute of Mechanism Theory, Machine Dynamics and Robotics
RWTH Aachen University
Eilfschornsteinstraße 18
52062 Aachen
(+49)-241 80-95546
akard@igmr.rwth-aachen.de www.igmr.rwth-aachen.de

You might also like