Development research in some industrial robotic control algorithms with many uncertain parameters

MINISTRY OF EDUCATION AND TRAINING VIET NAM ACADEMY OF SCIENCE AND TECHNOLOGY GRADUATE UNIVERSITY OF SCIENCE AND TECHNOLOGY ----------------------------- Nguyen Tien Kiem DEVELOPMENT RESEARCH IN SOME INDUSTRIAL ROBOTIC CONTROL ALGORITHMS WITH MANY UNCERTAIN PARAMETERS Major: Control Technique and Automation Code: 9.52.02.16 SUMMARY OF DOCTORAL DISSERTATION ON ELECTRONIC AND TELECOMMUNICATION TECHNIQUE Hanoi – 2018 The dissertation is completetd at Graduate University of Science and Technology and Vietnam Academy of Science and Technology. Scientific Instructor 1: Doctor Pham Minh Tuan Scientific Instructor 2: Doctor Nguyen Tran Hiep Review 1: Review 2: Review 3: . The dissertation is defensed at the Council of Doctoral Dissertation Evaluation - Graduate Universty level at Vietnam Academy Science and Technology at .. on date ..month .year 2011. The dissertation can be found at: - The library of Graduate University of Science and Technology - National Library of Vietnam INDEX PREFACE ................................................................................................ 1 CHAPTER 1. OVERVIEW ...................................................................... 2 1.1. Overview ........................................................................................... 2 1.2. Some applications for robot................................................................ 2 CHAPTER 2. SIMULATING PREFENRENTIAL MOTIONS FOR MOBILE ROBOTIC ARM AND DESIGNING OF SERVOING VISUAL FOLLOWING FLYING OBJECT. ..................................................................................... 3 2.2. Simulating preferential motions of camera on machine hand and designing of servo visual system following flying object. .......................... 3 2.2.1 Description of coordinates. ............................................................... 3 2.2.2. Preferentail motion. ......................................................................... 4 CHAPTER 3. SPEED CONTROL FOR ROBOTIC CAMERA SYSTEM FOLLOWING MOBILE TARGETS WITH MANY UNCERTAIN PARAMETERS........................................................................................ 8 3.2.1. Abstract. ......................................................................................... 8 3.2.2. Building control algorithms following mobile targets ...................... 8 3.2.3. Control algorithms visual servoing for pan/tilt base with many uncertain parameters. ................................................................................ 9 3.2.5. Conclusion on proposed control method. ....................................... 13 CHAPTER 4. CONTROL ALGORITHMS FOR INDUSTIAL ROBOTS USING ARTIFICIAL NEURAL NETWORK WITH ATTENTION TO THE ACTUATOR .................................................................................. 14 4.2.1 Control of robotic camera system following mobile targets with attention to the impact of the actuator . .................................................... 14 4.2.3. Control following mobile targets using neural network .................. 15 4.2.4. Result of simulating visual servo system with engine model/ simulation on Matlab. ............................................................................. 16 4.2.5. Conclusion on proposed control method ........................................ 18 CHAPTER 5. NON-LINEAR ADAPTIVE SLIDE CONTROL AGAINST DEGENERATION FOR ROBOTIC CAMERA WITH THE UNCERTAIN AND EXTERNAL NOISE MODEL ....................................................... 19 5.3. Kinetic model for robotic arm with fixed n-DOF .............................. 19 5.4. Designing control law ...................................................................... 19 5.6. Simulating control method. .............................................................. 21 5.7. Conclusion on proposed control method .......................................... 24 CONCLUSION OF THE DISSERTATION ........................................... 25 1 PREFACE 1. The necessity of the dissertation Controlling robot is still problematic due to the complexity, the nonlinearity and the uncertainty of the dynamical and kinetic equations caused by robots. Recently the controlling problem for robots with many uncertain parameters has received a lot of attention from researchers. Hence the researcher chooses the topic "Development research in some industrial robotic control algorithms with many uncertain parameters". 2. Research targets for the dissertation Proposing some control algorithms for robot-camera system following flying target. After doing research on some of control techniques of torque joints for the robot-camera system follows the mobile target and the robot-camera system with attention to the actuator following the mobile target. Finally, the author also proposes some control algorithms for robotic-camera arm system with uncertainty, external noise system against the degradation of the system, using nonlinear sliding mobile controller (TSMC) in combination with artificial neural networks to estimate uncertain parameters. 3. Some main contents of the dissertation - Develop an algorithm to control the robot-camera system mounted on a mobile platform following the target. - Develop an algorithm to control the robot-camera system in consideration the uncertain parameters using artificial neural networks with control signals for joints as torque signals. 2 - Develop an algorithm to control the robot-camera system including many uncertain parameters of the kinetic model and the mobile target with attention to the actuator using artificial neural networks. - Develop an algorithm to control the robot-camera system when there is uncertainty of the model and external noise using adaptive sliding control method and artificial neural network againts the degradation of the controller. CHAPTER 1. OVERVIEW 1.1. Overview Robots are used in many different areas such as simple turtle robots for teaching at high schools, welding robots in automobile factories, remote control robotic arms on the spaceship. Each application has its own problems, so the research field of robotics has actually emerged. There are many new emerging industries and many research results in this field while many more fields need to be explored and researched in the future along with many new perspectives that need to be developed and researched in laboratories. While people think that robotics is a unique field rather than a practical application, it is actually applied in manufacturing plants and received attention as well as application into production processes. 1.2 Some robotic application 1.2.1 Application in industrial 1.2.2 Applications in laboratory 1.2.3 Application in nuclear technology 1.2.4 Application in agriculture 1.2.5 Application in space exploring 1.2.6 Application in submersible survey equipment. 3 CHAPTER 2 SIMULATING PREFERENTIAL MOTION OF MOBILE ROBOTIC ARM AND DESIGNING NEW LAW OF VISUAL SERVOING FOLLING FLYING TARGET 2.2. Simulating preferential motion of camera on robotic arm and designing visual servo system following flying target 2.2.1 Description of coordinates Figure 2.5 One two-free-grade robotic arm mounted with camera on a mobile robot with wheel The homogeneous matrix shows the position and direction of OCXCYCZC in O0X0Y0Z0 given for the following formula: x y z px x x x5 534 34 34 x y z p0 y y y y5 534 34 34 C 0 x y z p5 5 z z z z 0 0 0 10 0 0 1 M M T s c s c c x xc c s s s c y yc c s h zc                              T (2.1) Z0 Y 0 X 0 O 0 Trục tilt Trục Pan Camera  4  5 Y C X C Z 3 X 3 Y 3 Z 4 Y 4 X 4 Z C 4 2.2.2. Preferential motion Jacobian matrix for robot has the following formula: 34 34 34 5 34 5 34 5 34 5 5 5 5 5 0 0 0 0 0 0 0 0 0 . 0 0 0 0 1 0 0 0 0 0 0 s c c s s s c c s c c c s s                      J (2.26) 2.2.3. Calculating the derivative of image characteristics Jacobi matrix of image characteristics 2 2 0 c c im 2 2 0 c c u uv u v z z v v uv u z z                      J We have the formula to calculate the derivative of image characteristics as follows . . im  ξ J J θ ζ (2.36) 2.2.4 Designing the Control Rule Figure 2.11- Diagram of proposed visual servoing controller following flying target Kinetic control (2.39) Kinetic controll (2.43) 2-DOF robotic arm with camera Chuyển động không xác định của Chuyển động không xác định của mục bay visual servoing folloing flying target + - - + v v d e  5 2.2.4.2 Kinetic Control Rule 4 5 4 5 v s u                ξ A ψ (2.37) ψ describes the variation of image characteristic deviation due to indefinite motion of the flying target. ψ can be estimated as following [16]: r 4r r 5 4r 5 ˆ - - - p e p e p e p e v s u                   A (2.38) With ψˆ is the estimated vector of ψ . Furthermore, r r 4, , p e p e  and r5 p e are the updated discrete data for ,ξ 4 , and 5 respectively. We can choose the desired angular velocity for pan-pilt joints as following: 14d 5d ˆA N n                      , (2.39) Replacing 4 5 T     in (2.37) by 4d 5d T     in (2.39), we have the following equation: 5 4d v n s u                N , (2.40) với ˆψ = ψ - ψ . 2.2.4.3 Kinetic Control Rule Kinetic model for pan/pilt platform/ base is shown as following: 6      ,q q v q  H h v v g , (2.41) (2.41) with  4 5 T  q , 4 5 T     v ,  4 5, T  τ , 4 is the torque at pan joint, 5 is the torque at tilt joint (see Figure 2.9). All  H q ,  ,h q v and  g q are shown specifically in simulating parameters. In order to design for kinetic control rule, the torque vector is chosen as following:      ,d d    τ Γe H q v h q v v g q , (2.43) with 4d 5d T d     v , d e v v . Γ is the constant matrix, positive diagonal line và can be chosen 2.2.5. Results of simulation Figure 2.13 a) The trajectory moves of the image characteristics in the image plane. b) Characteristics of coordinates per time. 0 2 4 6 x 10 -4 -4 -2 0 x 10 -4 truc U (m) tr uc V ( m ) 0 2 4 6 -1 -0.5 0 0.5 1 x 10 -3 thoi gian (s) to a do ( m ) u v quy dao dac trung anh huong cua chuyen dong 0 0.5 1 1.5 2 -3 -2 -1 0 1 thoi gian (s)S ai le ch v an to c go c (r ad /s )     7 Figure 2.14. Characteristics for e = v – vd per time Figure 2.15 Characteristics for the torque per time 2.2.6 . Conclusion on the proposed control method In this chapter, the author points out the process of simulating preferential motion of mobile robotic arm using Paul's algorithm. Then a new visual servoing rule for tracking flying targets is designed with the aim to make the image characteristic of the target asymptotic to the center of the image plane despite the trajectory of both flying objects and mobile robots are indefinite or unknown. Contrary to other control methods, visual servoing shows two strengths. Firstly, this method does not use the inverse pseudo matrix of the interaction matrix. Secondly, it also doesn't need to estimate the depth of the target. So the visual servoing method is more effective than other methods. The uniform stability of the whole system is ensured by Lyapunov standards. Simulation results with Matlab / Simulink software also confirm the accuracy and effectiveness of the proposed control method. (*) Main content of this chapter is published at scientific work no. [2]. Nguyen Tien Kiem, Hoang Thi Thuong, Nguyen Van Tinh, “Modeling the differential motion of a mobile manipulator and designing a new visual 0 0.5 1 1.5 2 2.5 3 3.5 4 -20 -10 0 10 thoi gian (s) M o m en q ua y (N .m )     8 servoing for tracking a flying target”, Informatics and control Journal - V.33, N.4 (2017), tr 339-355. CHAPTER 3. SPEED CONTROL FOR ROBOTIC CAMERA SYSTEM FOLLOWING MOBILE TARGET WITH MANY UNCERTAIN PARAMETERS. 3.2.1. Abstract Robotic camera system has two degrees of free rotation in two orientations: the azimuth (Pan) and the wrong angle (Tilt). This structure is widely used as a radar platform/base (fixed or mobile mounted on a vehicle or a ship) or a rotating platform for optical devices to monitor and check space. In this section, the author examines and studies the method of speed controlling of robotic joints with cameras mounted to mobile targets when we do not know the kinetic model for the platform/ base. Figure 3.2: Robotic camera system 3.2.2. Building control algorithms following mobile target e = M(ξ - ξ*) 9 [ , ]Tu vξ is image characteristic coordinate. The control purpose ensures if c o( (t)) *ξ r ξ , then e  0. To obtain this, we need to find the control rule based on image characteristics. From (3.4), control rule per camera velocity can be chosen as: 1 1 c c t      e Ω J e Jc (3.5) In which Jc -1 is the inverse matrix or pseudo inverse matrix (pseudo-inverse) for matrix Jacobi Jc. The equation now is (3.4) stable asymptotic in the shape of  e e . To stabilize the controller logarithmically, ee  ( > 0), we can choose the speed control rule for the camera as followings: c c t       e1 1J e JΩc (3.6) In which  is called degregation factor, the component t/ e is a characteristic component of the target's movement. Since the motion of the target is unknown, we must estimate the prediction during the control. Figure 3.4 Block diagram of robotic camera base speed control system 3.2.3. Visual servoing control algorithms for pan/tilt base when there are many uncertain parameters 10 When the robot model is unknown, it is not possible to choose the torque of the joints as (3.14). We can describe uncertain quantities in the Pan-Tilt pedestal dynamics in the form of ( ) ( ) ( ) ( ) ( ) ( ) q q q q q q     H H H h h h (3.18) In which q , qH( ) h( ) is the known part, q , q H( ) h( )are the unknown parts. Replacing (3.18) to (3.14) we have ( ) ( , )q q q q  H h f (3.19) with ( ) ( , )q q q f H q h (3.20) We choose the control torque τ with robotic joint as following 0 1    (3.21) 0 ( )( - ( - ) ( , ))d dq q q q q q H K h (3.22) In which d ε q q ; K is a positive symmetric matrix, which is the compensation control signal for uncertain components to be determined later. Replacing (3.21), (3.22) into (3.19) we have kinetic speed errors 1 1 1-ε +Kε = H (τ -f ) (3.23) Đặt 1 1-τ' = H τ (3.24) 1 1 -f' = H f (3.25) Replacing into (3.23) we have ' 'ε +Kε = τ -f (3.26) We will build a neural network with suitable algorithms to approximate 'f the network and determine the control signal 1τ so that the system (3.26) is stable asymptotic. 11 Theory 1: Robotic system Pan Tilt-camera 2 free grade with many uncertain parameters (3.19) with neuro network (3.28), (3.29) shall follow mobile target with error ( )d ε q - q 0 if we choose the control algorithms τ and the algorithms W for neuron network as following: 1q q,qd d τ = H( )(q -K(q-q )+h( )+ τ (3.30) 1 1         ε τ = H ( )Wσ - ε (3.31) T W εσ (3.32) In which free parameter K is the positive systematic matrix T K = K > 0 , with the parameters as , 0   Figure 3.6: The structure for visual servoing control camera system following mobile target 3.2.4. Simulation results of the visual sevoing control system on Matlab. -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 x 10 -3 -0.005 0 0.005 0.01 0.015 0.02 0.025 0.03 U axis V a x is dac trung anh 12 Figure 3.7 Diagram of image characteristics Figure 3.8 Diagram with the desired joint velocity Figure 3.9 Diagram of joint torque Figure 3.10 Diagram for neuron network weight 0 1 2 3 4 5 6 7 8 9 10 -1 0 1 2 3 4 Do Thi van toc goc khop mong muon cua Pan-Tilt Time (s) A n g u la r V e lo ci ty ( ra d /s ) q d dot 1 q d dot 2 0 1 2 3 4 5 6 7 8 9 10 -150 -100 -50 0 50 100 Time (s) M o m e n k h o p ( N .m ) Mo men khop Pan Mo men khop Tilt 0 1 2 3 4 5 6 7 8 9 10 -0.2 -0.1 0 0.1 0.2 0.3 do thi cac trong so mang no ron RBFNN thoi gian (s) ca c tr o n g s o w 1 w 2 w 3 w 4 13 Figure 3.11 Diagram of joint angles Figure 3.12 Diagram for errors in joint velocity 3.2.5. Conclusion on the proposed control method. In this chapter, the author has presented a method of building visual servoing system following the target. The simulation results on Matlab show that the algorithm given is convergence with high accuracy. Experimental studies on DPerception's actual use of robot models will be implemented in the near future. Research directions for the robot-camera system placed on mobile or ship vehicles are being studied with the help of inertial blocks in the problem of platform stability. 0 1 2 3 4 5 6 7 8 9 10 -1.5 -1 -0.5 0 0.5 1 1.5 do thi toa do cac khop thoi gian (s) to a d o k h o p ( ra d ) khop 1 khop 2 0 1 2 3 4 5 6 7 8 9 10 -4 -3 -2 -1 0 1 Time (s) S a i le c h v a n t o c k h o p P a n -T ilt e 1 =q 1 -q d1 e 2 =q 2 -q d2 14 (*) Main contents in this chapter shall be published at the scientific work no. [4] Nguyễn Tiến Kiệm, Pham Thuong Cat , „Velocity control for pan- tilt platform with camera following mobile target with uncertain parameters‟ , 6th Conference on mechatrocnics nationwide VCM2012, Hanoi dated on 14-15/12/2012, page 787-794. CHAPTER 4 CONTROL ALGORITHMS FOR INDUSTRIAL ROBOT USING ARTIFICIAL NEURON NETWORK WITH ATTENTION TO THE ACTUATOR 4.2 Controlling robotic camera system following mobile target with attention to the impact of the actuator The control task is performed through the difference function between the desired image characteristic const d ξ and the image characteristic obtained. This deviation function can be defined as follows: d e = (ξ - ξ ) (4.7) cx and xo respectively is the camera coordinates and target coordinates in the Cartesian coordinate system associated with the robot Control Rule Engine Ro bot Came ra Figure 4.5. Control diagram 15 platform. The kinetic equation of the robot is described by the following equation:  c x p q (4.8) The derivative per time (4.8), we get: c r t       p q x J q q The kinetic equation of robot and actuator are described as follows:     τ H q q h q,q (4.9) E E   Li Ri Kq t u (4.10) Tτ = K i       1 11 1 11 1 ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ( ) T E T T E                     K R u H GJ z H GJ GJq h K R Kq K R Li H G J q H GJ G Jq h R t (4.25)   11 ˆ ˆ T  ψ = RK H GJ (4.26)   1 1 1 1 ˆ ˆ( )T T E                 f RK H G J q H GJ G Jq h K t Li (4.27)   11 ˆ ˆ ˆ ˆ T        γ = RK -H GJ GJq h Kq (4.28) Combining the equations (4.25), (4.26), (4.27), (4.28), we obtain the following new equation: 1 E  ψz γ f u (4.29) 4.2.3. Control following mobile target using neuron network 16 0 1E  u u u (4.30)  0 D P   u ψ K z K z γ (4.31) u1 is the control signal to com