Jane Li. Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute

Transcription

1 Jane Li Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute

2 Use an example to explain what is admittance control? You may refer to exoskeleton What is ZMP, and how to determine it? You can draw a figure 2

3 3

4 In hand manipulation Coordinate many fingers and their independently controlled joints? Hand-arm coordination How to adapt the arm posture to maximize the benefit of hand? Bimanual coordination Temporal and spatial coordination? 4

5 Expressive morphology and behavior Interpreting human expression Natural response in physical human-robot interaction 5

6

7

8 Control architectures How to divide the control effort between a human and a robot? Communication channels Unilateral, bilateral 8

9 9

10 Passivity and stability Controllers? Many What make things worse Time delay Package loss 10

11 Teleoperation of mobile robots Kinematics dissimilarity? Driving an under-actuated robot? Collision avoidance? 11

12 Multi-lateral teleoperation 12

13 Haptics has all the problems that teleoperation has, and more I don t know what they will say to each other 13

14 Rendering virtual environment Virtual environment is discretized Stiff contact Complex geometry 14

15 Hardware development Haptic device is nothing but a robot Mechanical design, encoders and force sensors, actuation and transmission, You can turn a robot to haptic device, but remember High frequency for high perception fidelity Typical 1000 Hz 15

16 More than feeling a force You skin can feel a lot of things. How to render rich haptic perception? 16

17 Fidelity and richness of haptic perception On robot side interface development, sensors and sensor integration, On human side understanding the neuroscience of haptics, haptic perception stimulation Wearable haptics For daily activities Soft haptics and soft robotics

18

19 Building artificial humans to Understand humans Substitute humans Work with humans

20 Correspondence Demonstration interfaces Learning algorithms 20

21 Motion correspondence Perception correspondence

22 22

23 23

24 Demonstration data? Learning individual/compound motions? Learning task-specified motion/generalized motor skills? Learning from successes/failures? 24

25 Jane Li Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute

26 Lab equipment Course projects 26

27 Tele-nursing robotic system, under integration Vicon Mocap system Passive driver testbed 27

28 Available for projects Baxter robot Mobile base Compliant grippers 3D cameras 28

29 Available for projects Geomagic touch haptic devices Control software supporting various input devices 29

30 Hardware 10 cameras Vero 2.2 Max frame rate = 330 Hz Resolution 2048 x 1088 = 2.2 MP Covered space = 17 ft X 21 ft 8 on railing + 2 on ground (for closer view) Software Nexus human motion analysis Tracker object tracking 30

31 31

32 32

33 33

34 Vicon Nexus 2.5 documentation Vicon tutorial on you tube: OpenSim Documentation ntation 34

35

36 Human Motion Study (M) Mocap system, nexus motion analysis, OpenSim Hardware development (D) Mechatronics, CAD design, Java/C++ Programming Algorithm development (A) Mocap system, ROS, C++/Python, Matlab Virtual Reality and haptics (H) C++ programming Human robot interaction (I) Mocap system, C++/Python Programming 36

37 Study user experience in normal driving scenario Model integrated Controller under development Need human motion data to validate 37

38 Collection motion data from human subjects Steering wheel, shifting stick, pressing pedal Perform motion analysis to study Motion coordination, force responses 38

39 Hardware mostly ready Car racing simulator + Vicon Need a stick shift Need force sensors (from Greg?) Project contact Kene Mbanisi <kcmbanisi@wpi.edu> 39

40 Advantages Imitating motions of human characteristics Context-specific demonstration instances More intuitive to operators Challenges Collecting demos time-consuming and tedious Motor skills generalization 40

41 41

42 Teleoperation for reaching motion done Need to do Teleoperation for grasping control Algorithm for motion generalization Project contact: Sihui Li Srikanth Malla - only for setting up mocap teleoperation 42

43 43

44 Tasks Use 3D camera to tracking human hand position Reactive control robot receiver actively reaches out to take the object from human giver 44

45 Platform ready Humanoid nursing robot + kinect Need to do Setup vision-based control Implement human motion/intent prediction algorithm Project contact Gunnar Horve <gchorve@wpi.edu> 45

46 Home-based stroke rehabilitation Continuous, repetitive exercises at home Adapt to the patient s level of motor skills and training progress 3D printed haptic device Inexpensive Customized design Fast manufacturing 46

47 Specifications for development 3-DOF manipulator, 3D printed Motion range ~ 1 meter, to cover most of arm motion range End-effector payload ~ 4-6 kg Development support Kevin Harrington (harrington@wpi.edu ), robotics lab manager Reference: Quick Kinematic design Bowler Studio: Project Contact Nathaniel Goldfarb <nagoldfarb@wpi.edu> 47

48 48

49 Inter-arm coupling due to cross-hemisphere talk 49

50 Compare motion characteristics at joint and muscle levels Healthy subjects: Dominant vs Non-dominant arms Stroke patients: Healthy vs Paretic arms 50

51 Data has been collected and processed Need to apply machine learning algorithms to extract useful information 51

52 52

53 Arm Gross positioning Hand & fingers Precise reaching Fine manipulation 53

54 Coordinate manipulation endeffector with mobile manipulator Coordinate manipulation tool tip with surgical robot arm 54

55 Collect motion data on precise and dexterous manipulation Precise finger positioning & grasping, knob-turning, button-pressing Use machine learning algorithms to Generalize motor skills Study arm-hand-finger joint coupling 55

56 Prof. Cagdas Onal Motion control of prothetic limb Prof. Greg Fischer Developing hand exoskeleton 56

57 Fill in project selection form First and second choices for course project Preferred teammates Justification for your choice Previous course work, projects Reach out to your classmates actively Make sure your team members cover the required skills for the project 57