Some Contributions to Nonlinear Adaptive Control of PKMs

68
PhD Defense Thursday 17/12/2015 Some Contributions to Nonlinear Adaptive Control of PKMs From Design to Real - Time Experiments Antoine Ferreira Professeur Univ. d’Orléans Rapporteur Nacer Kouider M’sirdi Professeur LSIS, AMU, Marseille Rapporteur Mohamed Bouri Maître de conf. EPFL Examinateur Sébastien Krut CR CNRS LIRMM Examinateur François Pierrot DR CNRS LIRMM Direct. de thèse Ahmed Chemori CR CNRS LIRMM Co-encadrant Jury presented by: Moussab Bennehar

Transcript of Some Contributions to Nonlinear Adaptive Control of PKMs

Page 1: Some Contributions to Nonlinear Adaptive Control of PKMs

PhD Defense

Thursday 17/12/2015

Some Contributions to Nonlinear Adaptive Control of PKMsFrom Design to Real-Time Experiments

Antoine Ferreira Professeur Univ. d’Orléans RapporteurNacer Kouider M’sirdi Professeur LSIS, AMU, Marseille RapporteurMohamed Bouri Maître de conf. EPFL ExaminateurSébastien Krut CR CNRS LIRMM ExaminateurFrançois Pierrot DR CNRS LIRMM Direct. de thèseAhmed Chemori CR CNRS LIRMM Co-encadrant

Jury

presented by: Moussab Bennehar

Page 2: Some Contributions to Nonlinear Adaptive Control of PKMs
Page 3: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

• State of the art

• Dynamic modelling of parallel manipulators

• Proposed control solutions

Solution 1: Enhanced Model-Based Adaptive Control

Solution 2: Extended L1 Adaptive Control

How to deal with the internal forces issue ?

• Real-time experiments and results

• Conclusion and future work

Outline

Page 4: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Classification Non-adaptive Adaptive

Page 5: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Main existing control schemes for PKMs

Non-model-based Model-based Non-model-based Model-based

Non-adaptive schemes Adaptive schemes

PD with grav. comp. [Chifu10; Kelly97]

Computed torque and derivatives [Asgari15;

Sartori12; Zhang07]

Sliding mode [Jafarinasab11]

Predictive [Vivas03]

H-infinity [Becerra-

Vargas12; Rachedi15]

MRAC-based control schemes [Nguyen et al.,

1993]

Control schemes based on artificial networks [Li09]

Control schemes based on computed torque [Shang12]

Control schemes based on passivity [Honegger00;

Sartori15; Shang10]

PID control [Cheng03]

Nonlinear PID [Su04]

Fuzzy logic [Fang99;

Begon95]

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Classification Non-adaptive Adaptive

1

Page 6: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

No real-time adjustment of the parameters is required.Require relatively less computation time than adaptive schemes.Less control design parameters to be tuned.

May provide poor performance in the presence of large uncertainties.Large disturbances may lead to closed-loop instability.Cannot be used to estimate the parameters of the robot.Large uncertainties / disturbances may result in high-gain feedback.

Main existing control schemes for PKMs

Non-model-based Model-based

Non-adaptive schemes

PD with grav. comp. [Chifu10; Kelly97]

Computed torque and derivatives [Asgari15;

Sartori12; Zhang07]

Sliding mode [Jafarinasab11]

Predictive [Vivas03]

H-infinity [Becerra-

Vargas12; Rachedi15]

PID control [Cheng03]

Nonlinear PID [Su04]

Fuzzy logic [Fang99;

Begon95]

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Classification Non-adaptive Adaptive

2

Page 7: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Uncertainties and disturbances are estimated in real-time.Result in a better closed-loop performance if adequately tuned.Do not require accurate values of the robot’s parameters.

Require complex real-time computations.Large number of control design parameters (one parameter per estimation).Convergence of the parameters not always guaranteed.Estimated parameters may oscillate heavily leading to instabilities.

Main existing control schemes for PKMs

Non-model-based Model-based

Adaptive schemes

MRAC-based control schemes [Nguyen et al.,

1993]

Control schemes based on artificial networks [Li09]

Control schemes based on computed torque [Shang12]

Control schemes based on passivity [Honegger00;

Sartori15; Shang10]

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Classification Non-adaptive Adaptive

3

Page 8: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Dynamic Modeling of Parallel Robots

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Dynamic modeling Reformulation

Page 9: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Dynamic Modelling of PKMs

Simplifying hypotheses [Codourey98]:

Rotational inertia of the forearms neglected.

Mass of the forearms split up into two parts.

Friction effects are neglected.

Joint space inverse dynamic model

Inverse kinematics (IK)

Inverse differential kinematics (IDK)

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Dynamic modeling Reformulation

4

Page 10: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Reformulation of the Robot’s Dynamics for Adaptive Control

RegressorVector of parameters

to be estimated

Useful if all parameters are to

be estimated

If only a set of the parameters is to be

estimated Vector of known parameters

Vector of parameters to be estimated

Affine in the parameters formulation

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Dynamic modeling Reformulation

5

Page 11: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Proposed Control Solutions

State of the Art Modeling Proposed Solutions Experiments Conclusion

Overview Solution 1 Solution 2 Redundancy

Page 12: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed Control Solutions

Contribution 2

RISE-Based Adaptive Control

Contribution 1

DCAL with nonlinear

feedback gains

Contribution 4

L1 adaptive control with adaptive FF

Contribution 3

L1 adaptive control with nominal FF

Solution 1: Enhanced Model-Based Adaptive Control

Solution 2: Extended L1 Adaptive Control

Solu

tions

Cont

ribu

tions

Appl

icat

ions

Overview on proposed solutions

Overview Solution 1 Solution 2 Redundancy

6

Page 13: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Solution 1: Enhanced Model-Based Adaptive

Control

State of the Art Modeling Proposed Solutions Experiments Conclusion

Overview Solution 1 Solution 2 Redundancy

Page 14: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed Control Solutions

Contribution 2

RISE-Based Adaptive Control

Contribution 1

DCAL with nonlinear

feedback gains

Contribution 4

L1 adaptive control with adaptive FF

Contribution 3

L1 adaptive control with nominal FF

Solution 1: Enhanced Model-Based Adaptive Control

Solution 2: Extended L1 Adaptive Control

Solu

tions

Cont

ribu

tions

Appl

icat

ions

Overview on proposed solutions

Overview Solution 1 Solution 2 Redundancy

7

Page 15: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Desired Compensation Adaptation Law (DCAL) [Sadegh90]

Control law

Linear feedback term Auxiliary termModel-based adaptive feedforward

Parameters adaptation rule :

Overview Solution 1 Solution 2 Redundancy

8

Page 16: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed contribution

Main drawbacks

A feedback loop with constant gains.

Poor performance for nonlinear systems.

Sensitive to external disturbances.

Poor performance for high accelerations.

Limited tuning capabilities.

Desired Compensation Adaptation Law (DCAL) [Sadegh90]

Main advantages

Desired trajectories in the regressor.

No inversion of the mass matrix

required.

Reduced noise effect.

Reduced computing time.

Replace constant linear gains in the feedback loop by nonlinear ones

Overview Solution 1 Solution 2 Redundancy

9

Page 17: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

DCAL with nonlinear feedback gains [Bennehar14ab]

Nonlinear feedback gains

Large error large gain.Small error small gain.

Original control law

Proposed control law

[Bennehar14a] M. Bennehar, A. Chemori, and F. Pierrot. A New Extension of Direct Compensation Adaptive Control and Its Real-Time Application to Redundantly Actuated PKMs.

In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS14), pages 1670-1675, Chicago, Illinois, USA, September 2014.

[Bennehar14b] M. Bennehar, A. Chemori, and F. Pierrot. A New Revised Desired Compensation Adaptive Control for Enhanced Tracking: Application to RA-PKMs. Advanced

Robotics, 2015. [Submitted]

Error and combined error:

Overview Solution 1 Solution 2 Redundancy

10

Page 18: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Task-space trajectory generator

IK/IDKParallel robot

Model-based adaptive

feedforward

DCAL with nonlinear feedback gains [Bennehar14ab]

Overview Solution 1 Solution 2 Redundancy

11

Page 19: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed Control Solutions

Contribution 2

RISE-Based Adaptive Control

Contribution 1

DCAL with nonlinear

feedback gains

Contribution 4

L1 adaptive control with adaptive FF

Contribution 3

L1 adaptive control with nominal FF

Solution 1: Enhanced Model-Based Adaptive Control

Solution 2: Extended L1 Adaptive Control

Solu

tions

Cont

ribu

tions

Appl

icat

ions

Overview on proposed solutions

Overview Solution 1 Solution 2 Redundancy

12

Page 20: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Robust Integral of the Sign of the Error (RISE) [Xian04]

What is RISE ?

Robust Integral of the Sign of the Error

Non-model based feedback control strategy

Features a unique signum function

RISE control law

Overview Solution 1 Solution 2 Redundancy

13

Page 21: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed contribution

Main drawbacks

May lead to high-gain or high-frequency feedback.Poor performance in presence of large disturbances.Low performance in the case of hard nonlinearities

Main advantages

Stability of the system guaranteed. High order nonlinearities taken into account.MIMO systems supported.Large class of general disturbances assimilated.Very reasonable Hypotheses.

Augment RISE control with a model-based adaptive feedforward

Robust Integral of the Sign of the Error (RISE) [Xian04]

Overview Solution 1 Solution 2 Redundancy

14

Page 22: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

[Bennehar14c] M. Bennehar, A. Chemori, and F. Pierrot. A Novel RISE-Based Adaptive Feedforward Controller for Redundantly Actuated Parallel Manipulators. In IEEE/RSJ

International Conference on Intelligent Robots and Systems (IROS14), pages 2389-2394, Chicago, Illinois, USA, September 2014.

[Bennehar14d] M. Bennehar, A. Chemori, M. Bouri, L.F Jenni and F. Pierrot. A New Adaptive RISE-Based Control for Parallel Robots: Design, Stability Analysis and Experiments.

International Journal of Control. [Submitted]

RISE-based adaptive control [Bennehar14cd]

Proposed control law

Parameters adaptation rule

Model-based adaptive feedforward

Overview Solution 1 Solution 2 Redundancy

15

Page 23: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Task-space trajectory generator

IK/IDKParallel robot

Model-based adaptive

feedforward

RISE-based adaptive control [Bennehar14cd]

Overview Solution 1 Solution 2 Redundancy

16

Page 24: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed Control Solutions

Contribution 2

RISE-Based Adaptive Control

Contribution 1

DCAL with nonlinear

feedback gains

Contribution 4

L1 adaptive control with adaptive FF

Contribution 3

L1 adaptive control with nominal FF

Solution 1: Enhanced Model-Based Adaptive Control

Solution 2: Extended L1 Adaptive Control

Solu

tions

Cont

ribu

tions

Appl

icat

ions

Overview on proposed solutions

Overview Solution 1 Solution 2 Redundancy

17

Page 25: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Main Limitations of Conventional Adaptive Control

Overview Solution 1 Solution 2 Redundancy

High adaptation gain High gain/frequency feedback.1

System:

Adaptation gain:

18

Page 26: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Main Limitations of Conventional Adaptive Control

Overview Solution 1 Solution 2 Redundancy

High adaptation gain High gain/frequency feedback.2

System:

Adaptation gain:

19

Page 27: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Main Limitations of Conventional Adaptive Control

Solution

L1 Adaptive Control [Hovakimyan06]

Decoupled robustness and adaptation

High adaptation gain High gain/frequency feedback.

Adequate initialization of the parameters.

Persistence excitation of the parameters.

Specifications are specified only asymptotically.

Uncertainties may lie outside the actuators’ bandwidth.

1

2

3

4

5

Overview Solution 1 Solution 2 Redundancy

20

Page 28: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Background on L1 Adaptive Control [Hovakimyan06]

Inspired from direct Model Reference Adaptive Control (MRAC)

Overview Solution 1 Solution 2 Redundancy

21

Page 29: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

state predictor

low pass filter

projection operator

Background on L1 Adaptive Control [Hovakimyan06]

Inspired from direct Model Reference Adaptive Control (MRAC)

Overview Solution 1 Solution 2 Redundancy

22

Page 30: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Application of L1 Adaptive Control to PKMs [bennehar15e]

Tracking Error

Control Law

Adaptation Laws

Inverse dynamic model

Low-pass Filter Estimated Parameters

Error Dynamics

[bennehar15e] M. Bennehar, A. Chemori, and F. Pierrot. L 1 Adaptive Control of Parallel Kinematic Manipulators: Design and Real-Time Experiments. In IEEE International

Conference on Robotics and Automation (ICRA15), pages 1587-1592, Seattle, Washington, USA, May 2015.

Overview Solution 1 Solution 2 Redundancy

23

Page 31: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

state predictor

low pass filter

projection operator

Application of L1 Adaptive Control to PKMs [bennehar15e]

Overview Solution 1 Solution 2 Redundancy

24

Page 32: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed contribution

Main drawbacks

The estimated parameters lose their physical meaning.All uncertainties are considered with unknown structure.Do not take advantage of the knowledge about the dynamics.

Main advantages

Decoupled estimation and adaptation loops.

Specified performance ∀ 𝑡≥0.

Parameters boundedness.

No dynamic model is required.

Augment the L1 adaptive control with a model-based feedforward to further improve the tracking performance

Application of L1 Adaptive Control to PKMs [bennehar15e]

Overview Solution 1 Solution 2 Redundancy

25

Page 33: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Joint Space Inverse Dynamic Model

L1 Adaptive Control Law

Adaptation Laws

Proposed Control Law

Augmented L1 adaptive control with nominal model-based FF [bennehar15f]

State-feedback Term

Adaptive Term

Model-based feedforward

[bennehar15f] M. Bennehar, A. Chemori, and F. Pierrot. Feedforward Augmented L 1 Adaptive Controller for Parallel Kinematic Manipulators with Improved Tracking. IEEE

Robotics and Automation Letters (RA-L). [Submitted]

Overview Solution 1 Solution 2 Redundancy

26

Page 34: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Augmented L1 adaptive control with nominal model-based FF [bennehar15f]

state predictor

low pass filter

projection operator

Model-based feedforward

Overview Solution 1 Solution 2 Redundancy

27

Page 35: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed Control Solutions

Contribution 2

RISE-Based Adaptive Control

Contribution 1

DCAL with nonlinear

feedback gains

Contribution 4

L1 adaptive control with adaptive FF

Contribution 3

L1 adaptive control with nominal FF

Solution 1: Enhanced Model-Based Adaptive Control

Solution 2: Extended L1 Adaptive Control

Solu

tions

Cont

ribu

tions

Appl

icat

ions

Overview on proposed solutions

Overview Solution 1 Solution 2 Redundancy

28

Page 36: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Proposed contribution

Main drawbacks

Do not consider variations/uncertainties in

the model-based feedforward.

May lead to poor performance if the

dynamic model is not accurate.

Main advantages

Inherits the advantages of standard L1 adaptive control.Compensates for modeling nonlinearities and disturbances separately.May lead to better performance than standard L1-AC if accurate model.May reduce control effort.

Endow the additional feedforward term with adaptation capabilities

Augmented L1 adaptive control with nominal model-based FF [bennehar15f]

Overview Solution 1 Solution 2 Redundancy

29

Page 37: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Extended L1 adaptive control with adaptive model-based FF [bennehar15g]

Original L1 adaptive control law

Estimated uncertainties

Proposed extended L1 adaptive control law

Same as L1-ac

Model-based adaptive feedforward

Compensation of modeled uncertainties

[bennehar15g] M. Bennehar, A. Chemori, F. Pierrot and V. Creuze. Extended Model-Based Feedforward Compensation in L1 Adaptive Control for Mechanical Manipulators:

Design and Experiments. Frontiers in Robotics and AI.

State of the Art Modeling Proposed Solutions Experiments Conclusion

Overview Solution 1 Solution 2 Redundancy

30

Page 38: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

Task-space trajectory generator

Parallel robot

State predictor

Adaptation laws

Control law with low-pass filter

IK/IDK

Model-based adaptive

feedforward

Extended L1 adaptive control with adaptive model-based FF [bennehar15e]

Overview Solution 1 Solution 2 Redundancy

31

Model-based adaptive feedforward

Page 39: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Actuation Redundancy and Internal Forces

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Overview Solution 1 Solution 2 Redundancy

Page 40: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Experiments Conclusion

How to Deal with the Internal Forces Issue ?

2 DOFs and 3 actuators

Redundantly Actuated PKM

Redundantly actuated PKMs’ inputs contain antagonistic forces.

These forces create internal pre-stress.

They deteriorate performance, create vibrations and harm the robot.

These forces can be reduced using the projector [Muller11]:

The proposed control input becomes:

Identity matrix

Overview Solution 1 Solution 2 Redundancy

32

Page 41: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Real-Time Experiments and Results

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 42: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Experimental Platforms

Non-redundant Redundant

Veloce: 4 DOFs

Delta: 3 DOFs

Dual-V: 3 DOFs

Arrow: 4 DOFs

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

33

Page 43: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Non-redundant platforms

Delta robot Veloce robot

3 degrees of freedom (3T).

3 direct drive actuators.

20 Nm of maximum torque per actuator.

4 degrees of freedom (3T1R).

4 direct drive actuators.

127 Nm of max. torque per actuator.

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

34

Page 44: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Redundant platforms

Dual-V robot ARROW robot

3 degrees of freedom (3T).

4 direct drive actuators.

127 Nm of max. torque per actuator

4 degrees of freedom (3T1R).

6 linear actuators.

2500 N of max. force per actuator

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

35

Page 45: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Original DCAL vs DCAL with nonlinear feedback

gains

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 46: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Trac

king

err

ors

Nominal Case RobustnessEs

timat

ed P

aram

eter

s

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

36

Page 47: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

37

Page 48: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Original RISE vs Adaptive RISE

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 49: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Trac

king

err

ors

Nominal Case Robustness

Estim

ated

Par

amet

ers

RISE Adaptive RISERISE Adaptive RISE

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

38

Page 50: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

39

Page 51: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Original RISE vs Adaptive RISE

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 52: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Trac

king

err

ors

Estim

ated

Par

amet

ers

Nominal Case Robustness

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

40

Page 53: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

L1 Adaptive Control vs PD Control

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 54: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

41

Page 55: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

42

Page 56: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Original L1-AC vs Augmented L1-AC with nominal feedforward

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 57: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Trac

king

err

ors

Estim

ated

Par

amet

ers

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

43

Page 58: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

44

Page 59: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Original L1-AC vs Extended L1-AC with adaptive feedforward

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

Page 60: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Trac

king

err

ors

Estim

ated

Par

amet

ers

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

47

Page 61: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Platforms Results of Solution 1 Results of Solution 2

48

Page 62: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Conclusions and Future Work

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Conclusion Future Work Publuications

Page 63: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Tackled problemControl of parallel manipulators for high-speed trajectory tracking.

Main challengesComplex and highly nonlinear behavior.Uncertain / time-varying dynamics.Redundant actuation.

Proposed SolutionsSolution 1: Enhanced model-based adaptive controlSolution 2: Extended L1 adaptive control

ValidationReal-time experiments on available platforms:

Non redundant: Delta, Veloce.Redundant: Dual-V, Arrow.

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Conclusion Future Work Publications

49

Page 64: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

Investigate the use of other nonlinear gains for the proposed extended DCAL

controller.

The use of other adaptation terms and laws in combination of RISE control.

Implement and compare all developed controllers on one platform (Arrow for

instance).

Evaluate the performance of L1 adaptive control based methods for payload

changing scenarios.

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Conclusion Future Work Publications

50

Page 65: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

List of PublicationsJournal papers1. M. Bennehar, A. Chemori, M. Bouri, L.F Jenni and F. Pierrot. Adaptive RISE-based Control of the 3-DOFs Delta Parallel Robot. Submitted to International Journal of Control. 2. M. Bennehar, A. Chemori, and F. Pierrot. A New Revised Desired Compensation Adaptive Control for Enhanced Tracking: Application to RA-PKMs. Submitted to Advanced Robotics.3. M. Bennehar, A. Chemori, F. Pierrot and V. Creuze. Extended Model-Based Feedforward Compensation in L1 Adaptive Control for Mechanical Manipulators: Design and Experiments. Frontiers in Robotics and AI, 2015. To appear4. M. Bennehar, A. Chemori, S. Krut, and F. Pierrot. Control of Redundantly Actuated PKMs for Closed-Shape Trajectories Tracking with Real-Time Experiments. Transactions on Systems, Signals and Devices, 2015. To appear.

International conferences1. M. Bennehar, A. Chemori, and F. Pierrot. Feedforward AugmentedL1 Adaptive Controller for Parallel Kinematic Manipulators with Improved Tracking. Submitted to IEEE International Conference on Robotics and Automation (ICRA’16).2. M. Bennehar, A. Chemori, and F. Pierrot. L1 Adaptive Control of Parallel Kinematic Manipulators: Design and Real-Time Experiments. In IEEE International Conference on Robotics and Automation (ICRA’15), pages 1587-1592, Seattle, Washington, USA, May 2015.3. M. Bennehar, A. Chemori, and F. Pierrot. A Novel RISE-Based Adaptive Feedforward Controller for Redundantly Actuated Parallel Manipulators. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS’14), pages 2389-2394, Chicago, Illinois, USA, September 2014.4. M. Bennehar, A. Chemori, and F. Pierrot. A New Extension of Direct Compensation Adaptive Control and Its Real-Time Application to Redundantly Actuated PKMs. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS’14), pages 1670-1675, Chicago, Illinois, USA, September 2014.5. M. Bennehar, A. Chemori, S. Krut, and F. Pierrot. Continuous Closed-Form Trajectories Generation and Control of Redundantly Actuated Parallel Kinematic Manipulators. In Multi-Conference on Systems, Signals and Devices (SSD’14), Barcelona, Spain, 2014.

State of the Art Modeling Proposed Solutions Redundancy Experiments Conclusion

Conclusion Future Work Publications

51

Page 66: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

[Merlet06] J.P. Merlet. Parallel Robots, Second Edition. Springer, Dordrecht, Netherlands, 2006.

[Cheng03] Dynamics and control of redundantly actuated parallel manipulators. IEEE/ASME Transactions on Mechatronics, 8(4):483–491, December 2003.

[Codourey98] A. Codourey. Dynamic modeling of parallel robots for computed-torque control implementation. The International Journal of Robotics Research,

17:1325–1336, 1998.

[Su04] Y.X. Su, B.Y. Duan, and C.H. Zheng. Nonlinear pid control of a six-dof parallel manipulator. Control Theory and Applications, IEE Proceedings -, 151(1):95–

102, January 2004.

[Fang99] C. I-Fang, C. Hung-Hsiang, and L. Chin-Teng. Fuzzy control of a six-degree motion platform with stability analysis. In IEEE International Conference on

Systems, Man, and Cybernetics (SMC’99), pages 325–330, October 1999.

[Begon95] P. Begon, F. Pierrot, and P. Dauchez. Fuzzy sliding mode control of a fast parallel robot. In IEEE International Conference on Robotics and Automation

(ICRA’95), volume 1, pages 1178–1183, Nagoya, Japan, May 1995.

[Chifu10] Y. Chifu, H. Qitao, J. Hongzhou, P. O. Ogbobe, and H. Junwei. Pd control with gravity compensation for hydraulic 6-dof parallel manipulator. Mechanism

and Machine Theory, 45(4):666 – 677, 2010.

[Kelly97] R. Kelly. Pd control with desired gravity compensation of robotic manipulators a review. The International Journal of Robotics Research, 16(5):660–672,

1997.

[Asgari15] M. Asgari and M.A. Ardestani. Dynamics and improved computed torque control of a novel medical parallel manipulator: Applied to chest compressions

to assist in cardiopul monary resuscitation. Journal of Mechanics in Medicine and Biology, 15(04):1550051, 2015.

[Sartori12] G. Sartori Natal, A. Chemori, M. Michelin, and F. Pierrot. A dual-space feedforward pid control of redundantly actuated parallel manipulators with real-

time experiments. In IFAC Conference on Advances in PID Control (PID’12), March 2012.

[Zhang07] Y.X. Zhang, S. Cong, W.W. Shang, Z.X. Li, and S.L. Jiang. Modeling, identification and control of a redundant planar 2-dof parallel manipulator.

International Journal of Control Automation and Systems, 5(5):559–569, 2007.

[Jafarinasab11] M. Jafarinasab, M. Keshmiri, H. Azizan, and M. Danesh. Sliding mode control of a novel 6-dof parallel manipulator with rotary actuators. In

International Conference on Methods and Models in Automation and Robotics (MMAR’11), pages 218–223, August 2011.

[Vivas03] A. Vivas, P. Poignet, and F. Pierrot. Predictive functional control for a parallel robot. In IEEE/RSJ International Conference on Intelligent Robots and

Systems (IROS’03),volume3, pages 2785–2790, Las Vegas, Nevada, USA, October 2003.

[Becerra-Vargas12] M.Becerra-Vargas and E.M. Belo. Application of h-infinity theory to a 6dof flight simulator motion base. Journal of the Brazilian Society of

Mechanical Sciences and Engineering, 34(2):193–204, 2012.

[Rachedi15] M. Rachedi, B. Hemici, and M. Bouri. Design of an h controller for the delta robot: experimental results. Advanced Robotics, 29(18):1165–1181, 2015

References

Page 67: Some Contributions to Nonlinear Adaptive Control of PKMs

Moussab BenneharPhD defense

[Nguyen93] C. C. Nguyen, S. S. Antrazi, Z. L. Zhou, and C. E. Campbell. Adaptive control of a stewart platform-based manipulator. Journal of Robotic Systems,

10(5):657–687, 1993.

[Li09] Y. Li and Y. Wang. Trajectory tracking control of a redundantly actuated parallel robot using diagonal recurrent neural network. In International Conference

on Natural Computation (ICNC’09), volume 2, pages 292–296, Aug 2009.

[Shang12] W. Shang, S. Cong, and Y. Ge. Adaptive computed torque control for a parallel manipulator with redundant actuation. Robotica, 30(3):457–466, May

2012.

[Honegger00] M. Honegger, R. Brega, and G. Schweiter. Application of a nonlinear adaptive controller to a 6dof parallel manipulator. In IEEE International

Conference on Robotics and Automation (ICRA’00), volume2, pages1930–1935, San Francisco, CA, USA, April 2000.

[Sartori15] G. Sartori Natal, A. Chemori, and F. Pierrot. Dual-space control of extremely fast parallel manipulators: Payload changes and the 100g experiment.

IEEE Transactions on Control Systems Technology, 23(4):1520–1535, July 2015.

[Shang10] W. Shang and S. Cong. Nonlinear adaptive task space control for a 2-dof redundantly actuated parallel manipulator. Nonlinear Dynamics, 59(1-2):61–72,

2010.

[Sadegh90] N. Sadegh and R. Horowitz. Stability and robustness analysis of a class of adaptive controllers for robotic manipulators. The International Journal of

Robotics Research, 9(3):74–92, June 1990.

[Xian04] B. Xian, D.M. Dawson, M.S. de Queiroz, and J. Chen. A continuous asymptotic tracking control strategy for uncertain nonlinear systems. IEEE

Transactions on Automatic Control, 49(7):1206–1211, July 2004.

[Muller11] A. Muller and T. Hufnagel. A projection method for the elimination of contradicting control forces in redundantly actuated pkm. In IEEE International

Conference on Robotics and Automation (ICRA’11), pages 3218–3223, May 2011.

[Hovakimyan06] C. Chengyu and N. Hovakimyan. Design and analysis of a novel l1 adaptive controller, part i: Control signal and asymptotic stability. In American

Control Conference, pages 3397–3402, June 2006.

References

Page 68: Some Contributions to Nonlinear Adaptive Control of PKMs

Thank you !