1. Humanoid Robot Development of a simulation environment of an entertainment humanoid robot Lisboa-September-2007 Pedro Daniel Dinis Teodoro Orientador: Professor Miguel Afonso Dias de Ayala Botto Co-orientador: Professor Jorge Manuel Mateus Martins

2. 2 Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video GoalGoal State of Art Objectives This thesis was developed in collaboration with Robosavvy Ltd and boosted the creation of the Humanoid Robotics Laboratory of IDMEC-Center of Intelligent Systems, at Instituto Superior Técnico (http://humanoids.dem.ist.utl.pt/ ).http://humanoids.dem.ist.utl.pt/ The creation of the strong foundations for future developments in humanoid robots.

3. 3 GoalGoal Current commercially available humanoid robots are designed to perform motions using open-loop control. These robots are usually not able to move on uneven terrain and it is difficult or impossible to get them to perform movements that require instantaneous reaction to momentary instability. State of Art Objectives Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

4. 4 1.The establishment of a real-time protocol communication between the PC, using Matlab/Simulink® and the robot 2.The identification of internal and external properties of the humanoid robot. GoalGoal State of Art Objectives Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video 3.LQR implementation for stabilizing the humanoid robot on a high bar.

5. 5 The humanoid robot Hardware Architecture Software Architecture Bioloid humanoid robot from Robotis.com AX12+ Servo 1 MBps communication Speed. Full feedback on Position (300o), Speed, DC current, Voltage and Temperature. Can be set as an endless wheel. High Torque servos (1Nm). AX12+ Servo 1 MBps communication Speed. Full feedback on Position (300o), Speed, DC current, Voltage and Temperature. Can be set as an endless wheel. High Torque servos (1Nm). CM5 Controller the main controller of the humanoid. 57600 bps to receive/transmite data through servos and PC CM5 Controller the main controller of the humanoid. 57600 bps to receive/transmite data through servos and PC Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

6. 6 The humanoid robot Hardware Architecture Software Architecture Humanoid control architecture Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

7. 7 The humanoid robot Hardware Architecture Software Architecture C-MEX S-function written in C to communicate with the CM-5 throughout UART (universal asynchronous receiver / transmitter). C program for Atmega128 for completing the serial communication bridge. We have now a way to identify the parameters of the humanoid models making online experiments. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

8. 8 Mechanical Properties DC Servo Properties DC Servo Identification Mathematical Model Close Loop Pos Open Loop Speed Close Loop Pos Open Loop Speed Measured signals Measured signals Schematic of joint and link for two different humanoid configurations. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

9. 9 Mechanical Properties DC Servo Properties DC Servo Identification Mathematical Model Close Loop Pos Open Loop Speed Close Loop Pos Open Loop Speed Measured signals Measured signals Possible internal block diagram control of the servos. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

10. 10 Mechanical Properties DC Servo Properties DC Servo Identification Mathematical Model Close Loop Pos Open Loop Speed Close Loop Pos Open Loop Speed Measured signals Measured signals Experiments suggest that the servos do not have internally any angular velocity feedback control. Servos have an internal feedback position control loop Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

11. 11 Mechanical Properties DC Servo Properties DC Servo Identification Mathematical Model Close Loop Pos Open Loop Speed Close Loop Pos Open Loop Speed Measured signals Measured signals Dead-zone effect due to stiction. In our case this was clearly quantified to be around 7-10% of the full range. Experiments show that the output estimated velocity error is proportional to the voltage supplied to the servo. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

12. 12 Mechanical Properties DC Servo Properties DC Servo Identification Mathematical Model Close Loop Pos Open Loop Speed Close Loop Pos Open Loop Speed Measured signals Measured signals The dynamic characteristics of the servo are well captured by the BJ model. Box Jenkins (2,1,2,1) was found to best approximate the desired dynamical behavior of the servo. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

13. 13 The humanoid model SimMechanics simulator Virtual Reality animation The humanoid is treated as a three body serial chain in an inverted pendulum configuration. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

14. 14 The humanoid model SimMechanics simulator Virtual Reality animation The system is underactuated, being the motion of the legs and torso prescribed in order to stabilize the full body of the humanoid above the high bar. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

15. 15 The humanoid model SimMechanics simulator Virtual Reality animation Parent-Child hierarchy Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

16. 16 Equations of motion State-Space model Linear Quadratic Regulator The equations of motion for a generic n-link underactuated inverted pendulum deduced from the Euler-Lagrange equations. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

17. 17 Equations of motion State-Space model Linear Quadratic Regulator l (mm)lc (mm)m (g)I (gcm 2 ) Link 1 (Arms)143.668.7367.67890.7 Link 2 (Torso)115.857.5981.532898.6 Link 3 (Legs)184.0116.3576.411328.0 in order to use linear control algorithms, the system dynamics is linearized, using a first order Taylor's expansion at the vertical unstable equilibrium, q =[π/2,0,0] T and q =[0,0,0] T. State-space vector Physical properties Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

18. 18 Equations of motion State-Space model Linear Quadratic Regulator analyzing the zeros and poles of the system, it can be concluded that system is a non-minimum phase one Linear Quadratic Regulator was chosen, which provides a linear state feedback control law for the system Wothout Angle compensation Wothout Angle compensation With Angle compensation With Angle compensation Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

19. 19 Ideal servo servoIdeal Servo resolution Gyro resolution Servodead-zoneServodead-zone Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

20. 20 AchievedAchievedControlControl This project has successfully achieved the creation of the strong foundations for future developments in humanoid robots. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video Recommendations

21. 21 AchievedAchievedControlControl Recommendations LQR strategy was successfully applied in the stabilization of the humanoid on a high-bar although only in simulation and without the nonlinearities of the servos. Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

22. 22 AchievedAchievedControlControl Recommendations New gyroscope New nonlinear model containing the dynamics and nonlinearities of the servos New nonlinear controller (e.g. Sliding mode control) Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

23. 23 Guide Introduction Set Up Indentification Simulator Control Results Conclusions Video

24. Humanoid Robot Development of a simulation environment of an entertainment humanoid robot Lisboa-September-2007 Pedro Daniel Dinis Teodoro Orientador: Professor Miguel Afonso Dias de Ayala Botto Co-orientador: Professor Jorge Manuel Mateus Martins