autonomous systems lab 1 evaluation and optimization of rover locomotion performance machines that...

27
1 Autonomous Systems Lab Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07, Rome Workshop on Space Robotics

Upload: colleen-lorena-evans

Post on 11-Jan-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

1A

uto

nom

ou

s S

yste

ms L

ab

Evaluation and Optimization of Rover Locomotion Performance

Machines that know what they do

Thomas Thueer & Roland Siegwart

ICRA’07, RomeWorkshop on Space Robotics

Page 2: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

2

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Outline Locomotion Concepts Metrics Aspects Locomotion Performance Example: Rover Comparison - Simulation &

Hardware Improving Locomotion Performance Conclusion and Outlook

Page 3: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

3

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Locomotion Concepts How to design wheeled

rovers for rough terrain?

Page 4: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

4

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Characteristics of Locomotion Mechanisms

Trafficability: capacities to drive over a loose terrainMain parameters:

• Wheel-Ground Contact• Distribution of Mass

Maneuverability: mainly the steering capacitiesLocomotion mechanism (steering of wheels)Type of contact with ground

Terrainability: capacities to cross obstacles and maintain stability

Locomotion mechanismMass distributionType of contact, number and distribution of contact point

Page 5: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

8

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Wheeled Rovers (RCL):Concepts for Object Climbing

Purely frictionbased

Change of center of gravity(CoG)

Adapted suspension

mechanism with passive or active

joints

Page 6: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

9

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Catalog of Existing Solutions I

Page 7: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

10

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Catalog of Existing Solutions II

Page 8: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

11

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Metrics Necessary for proper comparison of different

systems “Know what conclusion you want to derive” Requirements

Precise definitionMeasurableObjectivity / independent from specific parameters Ideally available in simulation and reality

Apply to normalized systems Absolute / relative comparison Level of accuracy (requirements, level of

knowledge of final design)

Page 9: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

12

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Metrics – Overview Metrics for different aspects of performance

TerramechanicsObstacle negotiation capabilities

Metrics for sub-systemsEvaluation independently from roverSame performance of sub-system on different roversE.g. Rover Chassis Evaluation Tools (RCET) activity

for wheel characterization

Page 10: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

13

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

bMetricsTerramechanical & Geometrical Aspects

Analysis of wheel ground interaction based on Bekker

Drawbar pullEqual for all rovers if normalized, independent from suspension

Slope gradeabilityDepends on suspension that defines normal force distribution on slope

Static stabilitySee slope gradeabilityGeometrical analysis not sufficient!

Page 11: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

14

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

bMetricsObstacle Negotiation (Terrainability)

Minimum friction requirementMinimizing risk of slippage/getting stuck

in unknown terrainOptimization: equal friction coefficients

Minimum torque requirementMinimizing weight and power

consumption

SlipBad for odometry, loss of energy

Page 12: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

15

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison - Simulation & Hardware

Comparison of different roversCRAB (sim. & HW)RCL-E (sim. & HW)MER – rocker bogie type rover (sim.)

Page 13: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

16

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison –Simulation Setup

Performance Optimization Tool (2DS – RCET)Static, 2D analysisFast calculation allows for parametrical studies:

optimization of structuresOver actuated systems: optimization of wheel torquesResults reflect full potential of structure

(not influenced by parameter tuning, control algorithm)

SimulationsBenchmark: step obstacle (tough task for wheeled

rovers)Rovers normalized (mass, wheels, track, CoG, load dist.)Models with respect to breadboard dimensions/weight

Page 14: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

17

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison –Simulation Results

CRAB RCL-E MER

FWD

MER

BWD

Required

friction

coefficient [-]

0.64

0.95 0.57 1.0

Max. T [Nm] 6.0 7.3 6.7 8.9

Req

uir

ed

fri

cti

on

coeffi

cie

nt

[-]

Req

uir

ed

torq

ue

[Nm

]

Equally good performance of CRAB and MER Different forward and backward performance of

asymmetric systems as potential drawback

Page 15: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

18

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Rover Comparison – Experimental Setup

RoversModular design: same wheels

and electronicsGenoM software framework Motors: Maxon RE-max 22 Watt; EPOS controllersEqual footprint (0.65 m), similar weight (32-35 kg)

Test runsControl: velocity, velocity with wheel synchronizationTwo types of obstacle coating (rough, carpet-like)Step (wheel diameter high) At least 3 runs; log of currents, encoder values

Page 16: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

19

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison – Experimental Results (1)

CRABSuccess rate: SR = 100 %Slippage: Slip = 0.3 m

RCL-ESuccess rate: SR = 0 %

Wheels blocked because of insufficient torqueModification of controller settings: Maximum current

increased (2.5 A 3.5 A; 8.6 Nm 12 Nm)

Success rate: SR = 47 %Slippage: Slip = 0.41 m

Page 17: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

20

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison – Video of Testing

Hardware tests with CRAB and RCL-E

Page 18: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

21

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison – Experimental Results (2)

Rover: CRAB Successful test run Peaks indicate obstacle

climbing of wheels Current graph

Saturation at 2.5ANegative currents occur

Distance graph (encoders)Normal inclination

wheel moving or slipping

Reduced inclination wheel blocked

saturation

wheels blocked

negative currents

Page 19: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

22

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison – Experimental Results (3)

Rover: RCL-E Failed test: rover blocked

(current limit at 2.5 A)Rear wheel saturatedFront and middle wheel slip

Successful test(current limit at 3.5 A)

Current back wheel > 2.5 AFront and middle wheel:

currents similar as above Problems in climbing phase

can be detected (oscillation of signal)

wheels slipping

wheel slipping- lack of grip

Page 20: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

23

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Example: Rover Comparison –Simulation vs. Experiments

Qualitative AnalysisStrong correlation predictions – measurements Significantly higher torque (SR = 0 %, 2.5 A) and

friction coefficient (SR = 47 %, 3.5 A) of RCL-E than CRAB (SR = 100 %, 2.5 A)

Same ranking simulation/hardware for all metrics

Quantitative AnalysisDiscrepancy of numerical values (~40 %)Static, ideal model

Validation of simulations through hardware tests(Ref: Thueer, Krebs, Lamon & Siegwart, JFR Special Issue on Space Robotics, 3/2007)

Page 21: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

24

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Challenging Environment on Mars Spirit and Opportunity

Robots on Mars – since 24.1.2004

Page 22: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

25

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Motion Control – Tactile Wheels Improvement of locomotion

performance through motion control

Control typesTorque controlKinematics based velocity

control Need for tactile wheel

Wheel ground contact angle required

First prototype on OctopusDevelopment of new “metallic“ wheel

Page 23: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

26

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Flexible Wheels Better tractive performance Lower total motion resistance

Total sinkage [mm]

Wheel deflection

[mm]

Max. soil

slope [°]

Required wheel output torque [Nm]

Combined output power (6 wheels)

[W]

Required input power

[W]

Rigid wheel D=35 cm, b=15 cm, grouser height=3.4 cm, i=10 %

45.8 - 13.9

13.87 10.6 25.2

Flexible wheel D=35 cm, b=15 cm, grouser height=0.1 cm, pressure on rigid ground=5 kPa, i=10 %

12.9 12.8 13.9

6.17 4.7 11.2

Courtesy of DLR Köln

Page 24: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

27

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

bNavigation – Motion Estimation and Control in Rough Terrain

Page 25: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

28

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Conclusion Locomotion mechanisms and their characteristics Metrics for different aspects of performance Example of evaluation and comparison of systems

Focus on obstacle negotiation aspect of locomotion performance

Static 2D analysis in simulationVerfication and validation with hardware

How to improve performanceMotion controlTactile wheel as sensor for wheel ground contact angle

Page 26: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

29

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Outlook Continuous Flight on Mars

3.2 m

Page 27: Autonomous Systems Lab 1 Evaluation and Optimization of Rover Locomotion Performance Machines that know what they do Thomas Thueer & Roland Siegwart ICRA’07,

30

© Autonomous Systems Lab, ETH Zurich

Auto

nom

ous

Syst

em

s La

b

Thanks for your attention! Acknowledgement

This work was partially supported through the ESA ExoMars Program and conducted in collaboration with Oerlikon Space, DLR and vH&S

Questions ?