Manipulator Mechanical Design

1

1

MECH 498: Introduction to Robotics

Actuation, Sensing, and Design

M. O’Malley

2

Manipulator Mechanical Design• Particular structure of a manipulator influences kinematic and

dynamic analysis• The tasks that a manipulator can perform will also vary greatly with a

particular design (load capacity, workspace, speed, repeatability)

• The elements of a robotic system fall roughly into four categories– The manipulator mechanism & proprioceptive sensors– The end-effector or end of the arm tooling– External sensors (e.g. vision system) or effectors (e.g. part feeders)– The Controller

3

Manipulator Mechanical Design –Task Requirements

• Robots usually don’t fit the ideal of universally programmable devices

• Task Specific Design Criteria– Number of degrees of freedom– Workspace– Load capacity– Speed– Repeatability accuracy

4

Task Requirements - Number of DOF

• The number of DOF in a manipulator should match the number of DOF required by the task.

– Minimizes cost (hardware, computing power, and power consumption)

– Minimizes size/weight

5

Task Requirements

• Not all the tasks required 6 DOF for example:– End effector with an axis of symmetry - Orientation

around the axis of symmetry is a free variable,– Placing of components on a circuit board - 4 DOF

• Dividing the total number of DOF between a robot and an active positioning platform

6

Task Requirements• Workspace (Work volume, Work envelope)

– Placing in the work space of the manipulator– Singularities– Collisions

• Load Capacity– Size of the structural members– power transmission system– Actuators

• Speed– Robotic solution must compete on economic basis– Process limitations - Painting, Welding– Maximum end effector speed versus cycle time

• Repeatability & Accuracy– Matching robot accuracy to the task (painting - spray spot 8 +/-2 “)– Accuracy function of design and manufacturing (Tolerances)

2

7

Kinematic Configuration

• Joints & DOF -– For a serial kinematic linkages, the number of joints

equal the required number of DOF• Overall Structure

– Positioning structure (link twist 0 or +/- 90 Deg, 0 off sets)

– Orientation structure• Wrist

– The last n-3 joints orient the end effector– The rotation axes intersect at one point.

8

Kinematic Configuration - Cartesian

• Joints– Joint 1 - Prismatic– Joint 2 - Prismatic– Joint 3 - Prismatic

• Inverse Kinematics - Trivial• Structure -

– Stiff Structure -> Big Robot– Decoupled Joints - No singularities

• Disadvantage– All feeder and fixtures must lie “inside” the robot

9

Kinematic Configuration - Cartesian

10

Kinematic Configuration -Articulated

• Joints– Joint 1 - Revolute -Shoulder– Joint 2 - Revolute - Shoulder– Joint 3 - Revolute - Elbow

• Workspace– Minimal intrusion– Reaching into confined spaces– Cost effective for small workspace

• Examples– PUMA– MOTOMAN

11

Kinematic Configuration -Articulated

12

Kinematic Configuration - SCARA

• Joints– Joint 1 - Revolute– Joint 2 - Revolute– Joint 3 - Revolute– Joint 4 - Prismatic– Joints 1,2,3 - In plane

• Structure– Joint 1,2,3, do not support weight (manipulator or weight)– Link 0 (base) can house the actuators of joint 1 and 2

• Speed– High speed (10 m/s), 10 times faster then the most articulated

industrial robots• Example

– SCARA (Selective Compliant Assembly Robot Arm )

3

13

Kinematic Configuration - SCARA

14

Kinematic Configuration - Spherical

• Joints– Joint 1 - Revolute (Intersect with 2)– Joint 2 - Revolute (Intersect with 1)– Joint 3 - Prismatic

• Structure– The elbow joint is replaced with prismatic joint– Telescope

15

Kinematic Configuration - Spherical

16

Kinematic Configuration -Cylindrical

• Joints– Joint 1 - Revolute– Joint 2 - Prismatic– Joint 3 - Prismatic

17

Kinematic Configuration -Cylindrical

18

Kinematic Configuration - Wrist

• Joints– Three (or two) joints with orthogonal axes

• Workspace– Theoretically - Any orientation could be

achieved (Assuming no joint limits)– Practically - Severe joint angle limitations

• Kinematics– Closed form kinematic equations

4

19


• Three intersecting orthogonal Axes– Bevel Gears Wrist

• Limited Rotations

• Three Roll Wrist (Cincinatti Milacron)• Three intersecting non-orthogonal

axes• Continuous joint rotations (no limits)• Sets of orientations which are

impossible to reach

20


21


• Non intersecting axes wrist

• A closed form inverse kinematic solution may not exist

• Special Cases (Existing Solutions)– Articulated configuration

• Joint axes 2,3,4 are parallel– Cartesian configuration

• Joint axes 4,5,6 do not intersect 22

Actuation Schemes

23

Reduction & Transmission Schemes

24

Kinematic ConfigurationsDesign

• Decide degrees of freedom first• Then choose kinematic configuration to

obtain the best– Workspace– Dynamic properties– Use of actuators and sensors– Accuracy

• A general, 6 dof manipulator is usually classified by the first 3 dof plus a wrist

5

25

Workspace Attributes

• Design efficiency• How much material is needed to build different

designs with the same workspace?• Length sum

• Structural length index

• (W = workspace volume, di = distance between joint limits)

26

Condition of Workspace

• When the manipulator is near a singular point, actions of the manipulator are said to be poorly conditioned.

• Singular conditions are given by• Thus, use the Jacobian as a measure of

manipulator dexterity

27

Manipulability Measure (vel)

• Yoshikawa defines manipulability as

• For a nonredundant manipulator

• A good manipulator has a high w over large areas of its workspace

28

Manipulability (acc/force)• Asada examines the eigenvalues λI and

eigenvectors of the Cartesian mass matrix

• Graphically, this can be represented as an inertia ellipsoid

• This is the equation of an n-dimensional ellipsoid, where n is the dimension of X– Axis directions are eigenvectors and lengths are .– See Craig, figure 8.12.

29

Redundant structures

• Can be useful for avoiding collisions while operating in cluttered work environments

30

Closed Loop structures

• So far, we have only considered serial chain manipulators

• Closed loop, or parallel, structures can be stiffer and more precise

• But, they typically decrease joint ranges and therefore workspace size

6

31

6dof Parallel manipulator

• Stewart Platform– (inverse kinematics easy, forward kinematics hard!)

32

DOF for closed loop system

• DOF not readily obvious• Grubler’s formula for closed chain manipulators

states

– Where F is the total dof in the mechanism– l is the number of links (including the base)– n is the total number of joints– fi is the dof associated with the ith joint

• Stewart F = 6(14-18-1)+36 = 6

33

Actuators and Sensors

• First, choose general kinematic structure• Next, choose actuation

– Actuator– Reduction– Transmission

• Finally, select sensors• (And then control)

34

Actuator Location• Direct Drive

– Placed at the joint– Simple and high controllability– No transmission or reduction elements

• However, speed reduction is often required because many actuators are suited to high speeds and low torques

• Also, weight/inertia of actuators affect the dynamics, so the actuators are placed at or near the robot base. Thus, a transmission system must be used.

35

Reduction & Transmission

• Gears produce large reductions in a compact configuration

• Disadvantages:– backlash and friction

• Gear ratio: relationship between input and output speeds & torques

36

Types of gears• Spur Gears

– (parallel shafts)• Bevel gears

– (orthogonal shafts)• Worm gears/cross helical

gears– (skew shafts)

• Rack & Pinion• Consider load, wear and

frictionSRL, Georgia Tech

7

37

• Flexible Transmission– Bands, Cables, Belts

• Capstan drive used in haptic devices• Need large preloads to ensure the cable stays

engaged

Jake Abbott, JHU

38

Actuator Types

• Electric motors– DC (direct current)– Brushed– PM (permanent

magnet)• Pneumatic Actuators

39

PM DC brushed motors

• How do they work?– Rotating armature with coil

windings is caused to rotate relative to a permanent magnet

– current is transmitted through brushes to armature, and is constantly switched so that the armature magnetic field remains fixed.

40

DC Motor Components

41

DC motor terms

• Cogging– Tendency for torque output to ripple as the

brushes transfer power• Friction/damping

– Caused by bearings and eddy currents• Stall torque

– Max torque delivered by motor when operated continuously without cooling

42

Motor Equations

8

43

Pneumatic Actuators

• How do they work?– Compressed air pressure is used to transfer

energy from the power source to robotic device

• Many different types• Concerns are

– friction – bandwidth

44

Robot Sensors

• Manipulators– Proprioception, Force

• Mobile Robots– Dead reckoning, Tactile and proximity,

Ranging, etc.• Recommended reading:

– Mobile Robots by Joseph J. Jones and Anita M. Flynn

– Sensors for Mobile Robots by H.R. Everett

45

Manipulator Sensors

• Primary concern is proprioception• Kinesthesia/Proprioception/Force:

– A sense mediated by end organs located in muscles, tendons, and joints.

– Stimulated by bodily movements.

46

Potentiometers

• Produce a voltage proportional to shaft position

• Voltage divider

47

Potentiometers

• Problems:– Friction (for backdriveable systems like haptic

devices)– Noise– Resolution– Linearity

48

Optical Encoders• How do they work?• A focused beam of light aimed at

a matched photodetector is interrupted periodically by a coded pattern on a disk

• Produces a number of pulses per revolution (Lots of pulses = high cost)

• Quantization problems at low speeds

• Absolute vs. referential

9

49

Optical Encoders

• Phase-quadrature encoder– 2 channels, 90° out of phase– allows sensing of direction of rotation– 4-fold increase in resolution

50

Hall-Effect Sensors

• How do they work?– A small transverse voltage is generated across a

current-carrying conductor in the presence of a magnetic field

– (Discovery made in 1879, but not useful until the advent of semiconductor technology.)

51

Hall-Effect Sensors

• Amount of voltage output related to the strength of magnetic field passing through.

• Linear over small range of motion• Need to be calibrated• Affected by temperature, other magnetic objects in the

environments

52

Measuring Velocity

• Differentiate position– advantage: use same sensor as position

sensor– disadvantage: get noise signal

• Alternative– for encoders, measure time between ticks

53

Digital differentiation

• Many different methods• Simple Example:

– Average 20 readings = P1– Average next 20 readings = P2where t is the

the period of the servo loop• Differentiation increases noise!

54

Time-between-ticks

• Encoders fare poorly at slow velocities– There may be very few ticks during a single servo loop

• Instead, use a specialized chip (PLC) that measures time between ticks– Fares worse at high velocities

10

55

External sensors• Computer Vision

– Use vision to determine linkage position

• Magnetic– e.g., Ascension flock of

birds• Force

– Commercial load cells/force sensors

– Direct application of strain gages

56

Mobile Robot Sensing

• Transducing vs. understanding• Levels of abstraction:

– Is it light or dark?– Is there a wall to the left?– Who just walked in the room?

• Algorithms are required to determine the desired information from basic sensor data

57

Example

• Robart II– sonar, infrared,

bump, microwave motion, burglar alarm, surveillance camera, earthquake, & flood

58

Example

• Hannibal– Force, touch, color, potentiometer, force-sensing

whisker, gyroscope, pitch-and-roll, small camera, near-infrared rangefinder

59

Sensitivity

• Sensor output: r• Measured physical quantity: x• Sensitivity is S• Typical sensors output a voltage

corresponding to the value of x, so amplifiers can improve sensitivity

60

Range• Signals should be detected and amplified so that

the output falls in the correct range• Consider a signal where 0 = nothing and 255 =

maximum

11

61

More Sensor Metrics• Resolution:

– What is the smallest change you can measure?• Signal to Noise Ratio

– The signal is useful information and noise is anything else. A low ratio is bad!

62

Light Sensors

• Photoresistors (photocell)– Variable resistance, like a

potentiometer• Phototransistors

– Greater sensitivity• Photodiodes

– Highest sensitivity, but low output requires amplifier

63

Proximity Detectors• Near-infrared detectors

(IRs)– Signifies whether

something is present within a cone of detection

– Emitter-detector pair• Pyroelectric detector

– Output changes with small changes in temperature over time

– Detects radiation in the range of (8-10 µm)

– Useful for sensing of humans

64

Mobile Robot force sensors

• Microswitches for bumpers• Bend Sensors (conductive ink)

– 3-5x resistance change• Force-sensing resistors

– Several orders of magnitude resistance change

65

Cameras

• Video camera technology is rapidly changing

• CCD cameras can pick up near-infrared light

• On-board processing• Cell phone revolution

66

Sound sensors

• Microphones• Piezoelectric Film

– Really senses vibrations• Sonar

– Measure time-of-flight with emitter-detector pair (ping, then echo)

– Very commonly used in advanced robotics research

12

67

Use of SONAR data

68

Position and Orientation

• Encoders on Wheels– Dead reckoning– Does not account for slip

• Rate gyroscope– Determines speed of rotation

• Tilt sensors (e.g., mercury switch)• Compass

– ~45 degrees error due to metal components/indoors• GPS

69

Robot State

• Battery level– Time until recharging is required

• Stall current– Detects when wheels are not turning– Needs to respond slowly

• Temperature– Of motors, microprocessor

70

Commercial Mobile Robot

• Roomba vacuum cleaner from iRobot

Manipulator Mechanical Design

Documents