CONTROL METHOD FOR MOBILE PARALLEL MANIPULATORS

• US 20140188273A1
• Filed: 12/31/2012
• Published: 07/03/2014
• Est. Priority Date: 12/31/2012
• Status: Active Grant

• Current Assignee: King Fahd University Of Petroleum And Minerals
• Original Assignee: King Fahd University Of Petroleum And Minerals
• Inventors: KHOUKHI, AMAR, HAMDAN, MUTAZ M.
• Granted Patent: US 8,977,394 B2
• Time in Patent Office: Days
• US Class Current: 700/250
• CPC Class Codes: B25J 9/162 Mobile manipulator, movable base with manipulator arm mounted on it
• CPC Class Codes: B25J 9/1623 Parallel manipulator, Stewart platform, links are attached to a common base and to a common platform, plate which is moved parallel to the base
• CPC Class Codes: B25J 9/1664 characterised by motion, path, trajectory planning
• CPC Class Codes: G05B 2219/39073 Solve inverse kinematics by fuzzy algorithm
• CPC Class Codes: G05B 2219/40519 Motion, trajectory planning

Abstract

In the control method for mobile parallel manipulators, kinematic singularity and redundancy are solved through joint limits avoidance and manipulability criteria. By taking the MPM self-motion into consideration due to its redundancy, the inverse kinematic is derived using a hybrid neuro-fuzzy system, such as NeFIK. The discrete augmented Lagrangian (AL) technique is used to solve the highly nonlinear constrained multi-objective optimal control problem. An adaptive neuro-fuzzy inference system (ANFIS)-based structure (based on the result of the AL solution) is used to solve the online trajectory planning of the MPM.

Claims

1. A computer-implemented control method for a mobile parallel manipulator (MPM), comprising the steps of:

(a) setting trajectory parameters in an offline dataset, the trajectory parameters including initial and final positions, limits on state components, strut lengths, torques, accelerations, and sampling periods;

(b) formulating, off-line, a mathematical description of a constrained optimization problem for trajectory planning of the mobile parallel manipulator (MPM), the mathematical description including;

an energy and time cost minimization function and augmented Lagrange multipliers utilized in an augmented Lagrange decoupling procedure (ALD) to transform the problem into a non-constrained problem, the energy and time cost minimization function including inputs describing;

(i) a moving coordinate system related to self-movement of a base of the MPM relative to a fixed reference frame; and

(ii) torques and angular positions of actuators coupled to an end plate of the MPM;

an inertia matrix invertibility condition control law specification;

a discrete dynamic model approximation using a multi-step Adams-Bashforth predictive-corrective formulation;

a stopping criteria, wherein an iterative solution to the optimization problem describes at least one optimized trajectory;

(c) saving the at least one optimized trajectory;

(d) deriving at least one inverse kinematic solution based on the at least one optimized trajectory, wherein a joint position characterized by coordinates (q₁, q₂, q₃, q₄, q₅) is found from Cartesian coordinates (x, y, z, φ

);

(e) formulating, on-line, an optimized forward kinematic solution based on the at least one inverse kinematic solution;

(f) transferring to the MPM a resultant trajectory plan based on the optimized forward kinematic solution; and

(g) controlling the MPM actuators to constrain the MPM to execute motion consistent with the at least one optimized trajectory so that time, energy, and, alternatively, time and energy of the MPM motion is optimally minimized.

2. The computer-implemented control method according to claim 1, wherein said at least one inverse kinematic solution deriving step further comprises the step of training a Neuro-Fuzzy Inverse Kinematics (NeFIK) network utilizing back propagation to produce an error-minimized joint position in a preferred configuration, the NeFIK network being a multi-layer feed forward adaptive network having:

a first, two-input layer characterizing crisp values of the Cartesian position;

a last, three-output layer characterizing corresponding crisp joint values;

a first hidden fuzzification layer transferring the crisp inputs to linguistic variables via sigmoidal transfer functions;

a second hidden rule layer applying a Product t-norm to produce firing strengths of each rule; and

a third hidden normalization layer.

3. The computer-implemented control method according to claim 1, further comprising the step of training an Adaptive Neural Fuzzy Inference System (ANFIS) with said at least one inverse kinematic solution to further optimize said optimized forward kinematic solution.

4. The computer-implemented control method according to claim 3, wherein said further optimizing step comprises solely minimizing energy requirements of said forward kinematic solution.

5. The computer-implemented control method according to claim 3, wherein said further optimizing step comprises solely minimizing time requirements of said forward kinematic solution.

6. The computer-implemented control method according to claim 3, wherein said further optimizing step comprises minimizing time and energy requirements of said forward kinematic solution.

7. The computer-implemented control method according to claim 3, wherein said discrete dynamic model approximation using a multi-step Adams-Bashforth predictive-corrective formulation is characterized by the relations:

8. The computer-implemented control method according to claim 7, further comprising the step of characterizing said inertia matrix invertibility condition control law specification comprises the relation:

u= H(ζ

)v+ V(ζ

,{dot over (ζ

)})x₂+ G(ζ

).

9. The computer-implemented control method according to claim 8, further comprising the step of utilizing a Levenberg-Marquard version of the gradient back propagation algorithm as a training rule option in said NeFIK network while computing said energy and time cost minimization function.

10. The computer-implemented control method for a mobile parallel manipulator according to claim 9, further comprising the step of avoiding singularity and joint limits during said ALD calculation by computing an augmented function characterized by the relation:

{dot over (q)}_s={dot over (q)}_s1+{dot over (q)}_s2,

where {dot over (q)} represents vector joint rates of said MPM.

11. A computer software product, comprising a non-transitory medium readable by a processor, the non-transitory medium having stored thereon a set of instructions for performing a control method for a mobile parallel manipulator, the set of instructions including:

(a) a first sequence of instructions which, when executed by the processor, causes said processor to set trajectory parameters in an offline dataset, said trajectory parameters including initial and final positions, limits on state components, strut lengths, torques, accelerations, and sampling periods;

(b) a second sequence of instructions which, when executed by the processor, causes said processor to formulate, off-line, a mathematical description of a constrained optimization problem for trajectory planning of said mobile parallel manipulator (MPM), the mathematical description including;

an energy and time cost minimization function and augmented Lagrange multipliers utilized in an augmented Lagrange decoupling procedure (ALD) to transform the problem into a non-constrained problem, said energy and time cost minimization function including inputs describing (i) a moving coordinate system related to self-movement of a base of said MPM relative to a fixed reference frame and (ii) torques and angular positions of actuators coupled to an end plate of said MPM;

an inertia matrix invertibility condition control law specification;

a discrete dynamic model approximation using a multi-step Adams-Bashforth predictive-corrective formulation;

a stopping criteria, wherein an iterative solution to the optimization problem describes at least one optimized trajectory;

(c) a third sequence of instructions which, when executed by the processor, causes said processor to save the at least one optimized trajectory;

(d) a fourth sequence of instructions which, when executed by the processor, causes said processor to derive at least one inverse kinematic solution based on said at least one optimized trajectory wherein a joint position characterized by coordinates (q₁, q₂, q₃, q₄, q₅) is found from Cartesian coordinates (x, y, z, φ

);

(e) a fifth sequence of instructions which, when executed by the processor, causes said processor to formulate, on-line, an optimized forward kinematic solution based on said at least one inverse kinematic solution;

(f) a sixth sequence of instructions which, when executed by the processor, causes said processor to transfer to the MPM a resultant trajectory plan based on said optimized forward kinematic solution; and

(g) a seventh sequence of instructions which, when executed by the processor, causes said processor to control said MPM actuators, wherein said MPM executes motion consistent with the at least one optimized trajectory so that time, energy, and alternatively time and energy of the MPM motion is optimally minimized.

12. The computer software product according to claim 11, further comprising an eighth sequence of instructions which, when executed by the processor, causes said processor to train a Neuro-Fuzzy Inverse Kinematics (NeFIK) network utilizing back propagation to produce an error minimized said joint position in a preferred configuration, wherein said NeFIK is a multi-layer feed forward adaptive network having a first, two input layer characterizing crisp values of said Cartesian position, a last, three output layer characterizing corresponding crisp joint values, a first hidden fuzzification layer transferring said crisp inputs to linguistic variables via sigmoidal transfer functions, a second hidden rule layer applying a Product t-norm to produce firing strengths of each said rule, and a third hidden normalization layer.

13. The computer software product according to claim 11, further comprising a ninth sequence of instructions which, when executed by the processor, causes said processor to train an Adaptive Neural Fuzzy Inference System (ANFIS) with said at least one inverse kinematic solution to further optimize said optimized forward kinematic solution.

14. The computer software product according to claim 13, further comprising a tenth sequence of instructions which, when executed by the processor, causes said processor to solely minimize energy requirements of said forward kinematic solution.

15. The computer software product according to claim 13, further comprising an eleventh sequence of instructions which, when executed by the processor, causes said processor to solely minimize time requirements of said forward kinematic solution.

16. The computer software product according to claim 13, further comprising a twelfth sequence of instructions which, when executed by the processor, causes said processor to minimize time and energy requirements of said forward kinematic solution.

17. The computer software product according to claim 13, further comprising a thirteenth sequence of instructions which, when executed by the processor, causes said processor to, during said discrete dynamic model approximation, use a multi-step Adams-Bashforth predictive-corrective formulation is characterized by the relations:

18. The computer software product according to claim 17, further comprising a fourteenth sequence of instructions which, when executed by the processor, causes said processor to characterize said inertia matrix invertibility condition control law according to the relation:

u= H(ζ

)v+ V(ζ

,{dot over (ζ

)})x₂+ G(ζ

).

19. The computer software product according to claim 18, further comprising a fifteenth sequence of instructions which, when executed by the processor, causes said processor to utilize a Levenberg-Marquard version of the gradient back propagation algorithm as a training rule option in said NeFIK network while computing said energy and time cost minimization function.

20. The computer software product according to claim 19, further comprising a sixteenth sequence of instructions which, when executed by the processor, causes said processor to avoid singularity and joint limits during said ALD calculation by computing an augmented function characterized by the relation:

{dot over (q)}_s={dot over (q)}_s1+{dot over (q)}_s2,

where {dot over (q)} represents vector joint rates of said MPM.