Humanoid Robots 2: Dynamic Modeling modeling multi-body free - - PowerPoint PPT Presentation

humanoid robots 2
SMART_READER_LITE
LIVE PREVIEW

Humanoid Robots 2: Dynamic Modeling modeling multi-body free - - PowerPoint PPT Presentation

Autonomous and Mobile Robotics Prof. Giuseppe Oriolo Humanoid Robots 2: Dynamic Modeling modeling multi-body free floating complete model conceptual models for walking/balancing for running Oriolo: Autonomous and Mobile Robotics -


slide-1
SLIDE 1

Autonomous and Mobile Robotics

  • Prof. Giuseppe Oriolo

Humanoid Robots 2:

Dynamic Modeling

slide-2
SLIDE 2

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • multi-body free floating complete model
  • conceptual models

2

modeling

for walking/balancing for running

slide-3
SLIDE 3

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

3

like a manipulator?

can we consider this as a part (leg)

  • f a legged robot?

NO: this manipulator cannot fall because its base is clamped to the ground

this is a one-legged robot: Monopod from MIT

slide-4
SLIDE 4

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

4

floating-base model

the difference lies in the contact forces

  • ne may look at these contact configurations as different fixed-

base robots, each with a specific kinematic and dynamic model

  • r consider a single floating-base system with limbs that

may establish contacts

slide-5
SLIDE 5

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

the general model is that of a floating-base multi-body

floating-base model

5

slide-6
SLIDE 6

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

6

configuration vs.

formally similar

slide-7
SLIDE 7

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

dynamic equations (general form) but here we have a special structure where {g is the (Cartesian) gravity acceleration vector and Ji is the Jacobian matrix associated to the i-th contact force fi

7

Lagrangian dynamics

slide-8
SLIDE 8

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

joint torques only affect joint coordinates! to move x0 (i.e., the position of the reference body) the contact forces are necessary

8

Lagrangian dynamics

forces/torques accelerations mass/ inertia

  • centrifugal/Coriolis terms
  • joint torques
  • contact forces
slide-9
SLIDE 9

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

the second and third rows of the Lagrangian dynamics express the linear and rotational dynamics of the whole robot these correspond to the Newton-Euler equations, obtained by balancing forces and moments acting on the robot as a whole

9

Newton-Euler equations

slide-10
SLIDE 10

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

Newton equation: variation of linear momentum = force balance c : CoM position M : total mass of the system hence: we need contact forces to move the CoM in a direction different from that of gravity!

10

Newton-Euler equations

slide-11
SLIDE 11

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

Euler equation: variation of angular momentum = moment balance moments are computed wrt to a generic point o pi : position of the contact point of force fi L : angular momentum of the robot wrt its CoM

11

Newton-Euler equations

slide-12
SLIDE 12

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

recall: the moment of a force (or torque) is a measure of its tendency to cause a body to rotate about a specific point or axis angular momentum around the CoM: sum of the angular momentum of each robot link

12

Newton-Euler equations

fi pi - c pi c moment generated by the contact force fi around the CoM !k : angular velocity

  • f the k-th link
slide-13
SLIDE 13

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

in the equation of moment balance choose the point o so that is zero this is the Zero Moment Point (ZMP), i.e., the point wrt to which the moment of the contact forces is zero we denote this point by z

13

Zero Moment Point

slide-14
SLIDE 14

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

combine the Newton and Euler equations: divide the Euler equation by the z-component of Newton equation leads to

flat ground hypothesis (not necessarily horizontal)

14

Newton-Euler on flat ground

x z pi ground and we may have g

slide-15
SLIDE 15

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

15

Center of Pressure

the Center of Pressure (CoP) is a point defined for a set of forces acting on a flat surface flat ground: the CoP corresponds to the point of application of the Ground Reaction Force vector (GRF) note: GRF can also have a horizontal component (friction)

p

contact points

p1 p2 p3 p4

normal contact forces

slide-16
SLIDE 16

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

16

Center of Pressure

  • n flat ground, the moment balance equation tells us that the CoP

and the ZMP coincide the vertical component of the contact forces can only be positive (unilateral force) therefore the CoP/ZMP must belong to the convex hull of the contact points, i.e. the Support Polygon sufficient condition for balance

slide-17
SLIDE 17

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

flat ground first two components (x and y)

  • r in compact form

with

17

Newton-Euler on flat ground

slide-18
SLIDE 18

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

the Center of Pressure (CoP) z is usually defined as the point on the ground where the resultant of the ground reaction force acts we have 2 types of interaction forces at the foot/ground interface: normal forces fi

z and tangential forces fi x,y

the CoP may be defined as the point z where the resultant of the normal forces fi

z acts

the resultant of the tangential forces may be represented at z by a force f x,y and a moment Mt where ri is the vector from z to the point of application of fi

x,y

18

more on the CoP

slide-19
SLIDE 19

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

19

fi

z

f z z pi

x,y

z ri fi

x,y

f x,y Mt normal forces tangential forces f z z f x,y resulting GRF Mt the sum of the normal and tangential components gives the resulting GRF

more on the CoP

slide-20
SLIDE 20

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

we can analyze the effect of the various terms on the CoM horizontal acceleration (horizontal = in the x-y plane)

20

rewritten as

Lagrangian dynamics: multi-body system

slide-21
SLIDE 21

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

21

Lagrangian dynamics: multi-body system

aside from the effect of gravity (horizontal components) and variations of the angular momentum, the CoM horizontal acceleration is the result of a force pushing the CoM away from the CoP x-y plane support polygon

no gravity in x-y no variations of L

slide-22
SLIDE 22

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

22

Lagrangian dynamics: approximations

x z pi ground

  • n horizontal flat ground + CoM at constant height + neglect

c g

  • r

Linear Inverted Pendulum (LIP)

slide-23
SLIDE 23

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

23

Linear Inverted Pendulum interpretation

2 independent equations how the CoM moves in longitudinal direction (sagittal plane) lateral direction typical behaviors

slide-24
SLIDE 24

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

24

Linear Inverted Pendulum interpretation

px cx x cz px = zx in this case the ZMP zx coincides with the point of contact px of the fictitious leg z we can interpret the (longitudinal direction) LIP equation as a moment balance around px i.e.

  • Point foot

the simplest interpretation of the LIP is that of a telescoping (so to remain at a constant height) massless leg in contact with the ground at px (point of contact)

slide-25
SLIDE 25

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Point foot (longitudinal direction)

25

Linear Inverted Pendulum interpretation

px cx cz

step1 step2 t CoM

typical footsteps and CoM px = zx cx

may also be seen as a compass biped with only one leg touching the ground at the same time

slide-26
SLIDE 26

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Finite sized foot (with ankle torque ¿y)

since zx represents the ZMP location, there is no difficulty in extending the interpretation of the LIP considering both single and double support phases with a finite foot dimension

26

single support double support

zx zx

support polygon

Linear Inverted Pendulum interpretation

slide-27
SLIDE 27

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Finite sized foot (with ankle torque ¿y)

we can see the single support phase from the stance foot point of view i.e. with the dynamics of the rest of the humanoid represented by an equivalent fictitious leg. A way to keep the CoM balanced is using an equivalent ankle torque (the real joint torques are such that an equivalent ankle torque is applied)

27

zx

Linear Inverted Pendulum interpretation

CoP

¿y

ankle torque finite foot

px

finite foot with equivalent massless leg plus ankle torque CoP

+

¿y ¿y

moment w.r.t. zx (CoP) = 0

slide-28
SLIDE 28

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Finite sized foot (with ankle torque ¿y)

note: it is possible to move the CoP through the ankle torque ¿y without stepping

28

single support

zx

Linear Inverted Pendulum interpretation

CoP

¿y

ankle torque finite foot

px

  • CoP = -zx

finite foot with equivalent massless leg plus ankle torque

slide-29
SLIDE 29

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Finite sized foot (with ankle torque ¿y)

29

longitudinal direction

typical footsteps with single and double support: for example, in the first single support (- -) the left foot is swinging; as soon as the right foot touches the ground the double support starts (—) and the ZMP moves from the left to the right foot (longitudinal and lateral motions)

Linear Inverted Pendulum interpretation

SS: single support DS: double support

with

slide-30
SLIDE 30

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

30

zx

support polygon

Linear Inverted Pendulum interpretation

zx

finite foot with equivalent massless leg plus ankle torque and reaction mass

¿y ¿wy J

  • Finite sized foot and reaction mass

it is also possible to extend the point-mass to be a rigid body with its rotational inertia so that also a hip movement can be modelled

slide-31
SLIDE 31

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Finite sized foot and reaction mass

what is the effect of the rotating inertia around the CoM?

31

Linear Inverted Pendulum interpretation

px px ¿wy GRF GRF a reaction mass type pendulum, by virtue of its non-zero rotational inertia, allows the ground reaction force to deviate from the lean

  • line. This has important implication in gait and balance.

since no rotational inertia around the CoM, a different direction of the GRF would create a non-zero moment

slide-32
SLIDE 32

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

  • Finite sized foot and reaction mass

moment around ankle

32

Linear Inverted Pendulum interpretation

zx ¿y ¿wy J GRF i.e. with ankle torque moves the CoP while the reaction mass torque changes the GRF direction

to highlight the presence of a non-zero moment around the CoM a new point named Centroidal Moment Pivot (CMP) is introduced and defined as the point where a line parallel to the ground reaction force, passing through the CoM, intersects with the external contact surface

  • CMP

CMP

slide-33
SLIDE 33

Oriolo: Autonomous and Mobile Robotics - Humanoid Robots 2

Although extremely simplified, the LIP equation describes in first approximation the time evolution of the CoM trajectory. Moreover

  • it defines a differential relationship between the CoM trajectory and

the ZMP (or CMP) time evolution

  • it is easier to design a controller which makes the actual CoM follow a

desired behaviour

  • dynamic balancing will be characterized in terms of the ZMP
  • the problem will then be to understand which CoM trajectory,

solution of the LIP equation, guarantees that dynamic balancing is achieved

33

Linear Inverted Pendulum: basic scope