MIT Confidential

High-Accuracy, Repeatable Wrist Interface

Pat Willoughby (pjwst10@mit.edu)Prof. Alexander Slocum, Advisor

MIT Precision Engineering Research Group

August 15, 2001

MIT Confidential

Overview of Common Coupling Methods

Elastic Averaging

Non-Deterministic

Pinned Joints

No Unique Position

Kinematic Couplings

Kinematic Constraint

Flexural Kin. Couplings

Kinematic Constraint

Quasi-Kinematic Couplings

Near Kinematic Constraint

MIT Confidential

Exact Constraint (Kinematic) Design• Exact constraint means a component has an equal number of constrained points to

number of degrees of freedom

• If component is over constrained, clearance and high tolerances required to prevent premature failure or assembly incompatibility

• Kinematic design means that the motion is exactly constrained and geometric equations can be written to describe its motion

• Kinematic Couplings constrain components exactly, commonly providing repeatability of ¼ micron or on the order of parts’ surface finish

• Managing Hertz contact stresses is the key to successful kinematic coupling design

MIT Confidential

Prototype Coupling Designs – Canoe Ball

MIT Confidential

Prototype Coupling Designs – Three Pin

MIT Confidential

Schematic of Three Pin Coupling Design

Two “Pins” on Arm Plate

Two “Holes” On Wrist Plate

Preload Bolt and Third Pin

MIT Confidential

Prototype Wrist Plate MountingTests at ABB Robotics Vasteras, July/August 2001: Test static (bolted) and dynamic

(5-point path) repeatability of canoe ball and three-pin wrist prototypes

Test variety of preloads (canoe balls)

Replacement in two orientations (45 and 90 degrees to ground)

Measure tool point motion using Leica LTD500 Laser Tracker

Repeatability of robot path + measurement system approximately 20 microns

Average Repeatability for Each Case

0.0000

0.0500

0.1000

0.1500

0.2000

0.2500

0.3000

0.3500

0.4000

0.4500

Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9 Case 10 Case 11 Case 12 Case 13 Case 14

Average Repeatability (mm)

MIT Confidential

Repeatability Performance

Canoe balls vs. Normal Wrist @ 45 deg = 35% reduction

Canoe balls vs. Normal Wrist @ 90 deg = 64% reduction

Three-pin vs. Normal Wrist @ 45 deg = 44% reduction

Repeatabilty for Best Cases

0.0000

0.0200

0.0400

0.0600

0.0800

0.1000

0.1200

Normal Wrist Canoe Balls withProper Preloading,

Static 45 degPosition

Canoe Balls withProper Preloading,Dynamic 5 Point

Positions

Three Pin, BeforeDamage, Dynamic5 Point Positions

Canoe Balls withProper Preloading,

Static 90 degPosition

Re

pe

ata

bil

ity

(m

m)

MIT Confidential

Recommended Next Steps Test three pin coupling in lab setting

for ideal case repeatability

Adapt canoe ball design to fit into space of wrist

Suggest production designs for different concepts

Investigate:

• Three pin coupling in 90 degree position

• Effect of friction reduction using TiN coated elements or lubrication

• Coupling design independent of mounting position

• Applicability quasi-kinematic couplings

Evaluate long-term dynamic performance

MIT Confidential

High-Accuracy, Quick-Change, Robot Factory Interface

John Hart (ajhart@mit.edu)Prof. Alexander Slocum, Advisor

MIT Precision Engineering Research Group

August 15, 2001

MIT Confidential

Design, test, and demonstrate production feasibility of a modular robot baseplate with kinematic couplings as locators: A repeatable, rapidly exchangeable interface

between the foot (three balls/contactors) and floor plate (three grooves/targets)

Calibrate robots at ABB to a master baseplate Install production baseplates at the customer

site and calibrated the kinematic couplings directly to in-cell tooling

Install robot according to refined mounting process with gradual, patterned preload to mounting bolts

TCP-to-tooling relationship is a deterministic frame transformation

Base calibration data handling is merged with ABB software, enabling 0.1 mm TCP error contribution from repeatability and exchangeability error of kinematic couplings

Project Goals

MIT Confidential

Design 3-point kinematic coupling mounts for the 6400R foot:

Prototype Coupling Designs

Canoe Ball Six “point” contacts 0.5 m radius ball surface 20 mm diameter elastic

Hertzian contact

Three-Pin Three line + three surface

contacts In-plane preload overcomes

friction to deterministically seat pins

Vertical bolt preload engages horizontal contact surfaces

MIT Confidential

Prototype Coupling Designs

Groove/Cylinder Twelve line contacts Aluminum cylinders Apply bolt preload

(elastic deflection of cylinders) for dynamic stability

MIT Confidential

Prototype Base MountingTests at ABB Robotics Vasteras, July/August 2001: Static (bolted) and dynamic (5-

point path) repeatability of canoe ball and three-pin interfaces

Static (manipulator rest only) repeatability of groove/cylinder interface

Test both basic (air wrench) and refined (torque wrench, greased bolts) mounting processes

Measure tool point motion using Leica LTD500 Laser Tracker

Repeatability of robot path + measurement system approximately 20 microns

MIT Confidential

Repeatability Performance

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

BMWbase (diff.

robots)

Three-pin:basic

mounting

Three-pin:refined

mounting

Canoeballs:basic

mounting

Canoeballs:

refinedmounting

Aluminumcylinders:

(static)basic

mounting

Aluminumcylinders:

(static)refined

mountingDesign

Rep

eata

bilit

y [m

m]

Three-pin vs. BMW base = 83% reduction

Canoe balls vs. BMW base = 85% reduction

Cylinders vs. BMW base = 92% reduction

Refined mounting vs. basic mounting = 50-70% reduction

8-bolt blue pallet repeatability (not shown) = 1.63 mm

MIT Confidential

Exchangeability PerformanceSimulate exchangeability error from manufacturing variation: Calibrate interfaces by

measuring contacts and calculating interface error transformation

Model direct measurement of pins + contacts, and offset measurement of canoe balls

Exchangeability is error between calculated and true interface transformation, given chosen level of calibration and manufacturing tolerances (low, med, high)

250-trial Monte Carlo simulation in MATLAB at each calibration level

Three-pin exchangeability:

0 = no interface calibration3 = full (x,y,z) of pins and

contact surfaces

MIT Confidential

Total Mechanical Accuracy“Quick-Change” Accuracy = Repeatability +

Exchangeability

Interface calibration decouples accuracy from manufacturing tolerances of mounting plates and couplings (if direct measurement of contacts)

Results show repeatability is highly f(mounting process) – this may present a performance limit for factory mountings; interface should be micron-repeatable under perfect conditions

Totally, a near-deterministic prediction of robot interface accuracy

Canoe ballsThree-pinGroove/cylinder

(measured) (simulated)

0.22 mm = 0.06 + 0.16*

0.12 mm = 0.07 + 0.05 - = 0.03** + (Incomplete)

*driven by error of offset position measurement**static only

MIT Confidential

Recommended Next Steps Test groove/cylinder interface with

preload + motion

Test traditional quasi-kinematic couplings

Evaluate long-term dynamic performance

Design production three-pin adaptation to BMW 2-pin base

Evaluate canoe ball 4-point mounting for Voyager?

Build kinematic coupling “Expert System” – combine test results, simulation results, etc. into design tool that gives minimum cost design recommendation and tolerance budget as f(accuracy requirement)