Imperial College London by yy8Shk


									My research and teaching

Speaker: Chenguang Yang
Short biography

09/2001-07/2005, B. E. in measurement and control
College of Automation
Northwestern Polytechnical University (NPU), China.

08/2005-09/2009, Ph. D in adaptive/intelligent control
Department of Electrical & Computer Engineering &
Social Robotics Lab,
National University of Singapore (NUS), Singapore

10/2009-presesent, Research Associate
Human Robotics Group
Department of Bioengineering
Imperial College London, UK
       Summary of research experience
control &
                                                     Mobile robot localization and navigation

                                                                Mobile manipulator control
Bringing Human Neuromotor Intelligence into Robots
        Contribution to control theory I

 Prediction based control

       Unknown vector parameters.

                                                                               External disturbance

                                                                      Unknown control gains

We developed controller based on future states prediction (as well as with estimation of the uncertain
systems) to solve non-causal problems (Ge, Yang &Lee, 2008)
        Contribution to control theory II

 Overcome of the unknown control direction problem

The sign of coefficient of control input, known as control direction, decides along which direction the
control operates, and is very important in the parameter estimation.

 We solved this problem with exploit of Nussbaum gain (defined on a sequnce xs) with above
 oscillation properties (Yang, Ge & Lee, 2009)
        Contribution to control theory III

 Compensation of uncertain nonlinearities

The effect of uncertain nonlinearities should be compensated for stability and asymptotic tracking

We introduced the time index                                             Uncertain nonlinearities

which uses previous states information to compensate for uncertainties at current instant according to

(Ge, Yang, Dai & Lee, 2009)
        Contribution to control theory IV

 Lift approach for periodic varying parameters

When a parameter is time varying with period N, then we introduce an augmented parameter

and an augmented state

so the constant parameter estimation can be applied (Yang, Ge & Li, 2010)
        Contribution to control theory V

Implicit control approach

It is not easy to design a controller for the general nonaffine systems, in which control input appear in
an nonaffine manner so that the feedback linearization approach is not applicable.

We then use implicit function theorem to identify an inverse implicit control. Neural network is used to
online emulate the implicit control with guaranteed stability (Yang et al, 2009)
        Contribution to control theory VI

 Hidden leader following adaptive control of multi-agent system

For every agent with dynamics

and influenced by agents from neighbors by

We developed an adaptive control for each agent to track a target through following a hidden leader,
which is unknown to other agents (Ge, Yang, Li & Lee, 2008)
        Mobile robot control
 Mobile robotic manipulator control

We have applied adaptive and neural network control techniques to force/motion control of mobile
robot. The mobile manipulator is controlled to track prescribed trajectory along the contact surface
while maintain a constant force normal to the surface. (Li, Yang & Gu, 2007; Li, Yang &Wang, 2007)

•Adaptive and neural                                                   Constraints
network approach:
compensation for
internal uncertainties

•Robust control
approach:                                                         Non-
                                                                 自然拘束                    人工拘束
                                                                holonomic               Holonomic
compensation for
external disturbance.                                           mobile                   contact
                                                                platform                 surface
        Multi-robot cooperation
 Mobile robotic manipulator control

We have further studied multiple (two or more) mobile manipulators cooperation. The task is to grasp a
common object in contact with a surface. (Li and Yang, 2010, submitted)

 •Maintain internal force to hold
 the object.

 •Track the prescribed reference
 trajctory on the surface.

 •Guarantee the interaction force
 between object an surface

Decentralized dynamics of the whole system is
developed and then decentralized control is
designed using robust and adaptive approach.
       Autonomous robot project
Our X-1 team for the TechX challenge, Singapore, 2008

                                                             Zhang, Tao, Abraham, Rebsamen,
                                                             Yang and Ge, 2010, MFI

The TechX Challenge Urban Mobile Robot Competition, similar to the DARPA Grand Challenge of
US, was organized by the Defense Science and Technology Agency of Singapore.
Overall system Structure
              My contributions
Positioning and navigation

• KF based SLAM algorithm

•Corner detection as lardmark

•Laser scanner based navigation

•Topologic map building
                                  Corner detection         Laser scan plot

                                         Incremental map
                                  Stairs climbing
             Elevator operation

                                       Indoor navigation
Video demo
        Human impedance control

Hogan in1985
hypothesized that our
CNS is able to adapts
endpoint force and
independently for
different task
                        Impedance control in humans was proved in experiment study by Burdet
                        et al. published in nature [Burdet et al. 2001]. Some empirical models are
                        built from then and a strict modeling using techniques from control
                        theory is expected.
Sensor reference trajectory adaptation
Bringing human neuromotor intelligence into robots I
Bringing Human Neuromotor Intelligence into Robots I
Predictive adaptive control
Predictive adaptive control
        Teaching experience at NUS

 Several control and robotics courses

•Project design and marking

•Lab demonstration

•Tutorial preparation

•Homework assignment and marking
        Teaching experience at Imperial College London

 Neuromechanical control and learning

•Lecture on the part of neuromuscular skeleton
kinematics and dynamics

•Tutorial assignment and marking

•Project design and marking

•Help on the preparation of exam paper
        Supervision experience

 Final year project of undergraduates:

•Koh, Chin Chao, Adaptive control of nonlinear systems, National University of Singapore, 2006
•Wang, Huidong, Virtual robotic control platform, National University of Singapore, 2007
•Zhang, Yang, Servo control of low cost robotic fingers, National University of Singapore, 2008
•Li, Wanning, Internet based remote control system, National University of Singapore, 2009
•Wang, Tianyi, Developing an algorithm to find the optimal stiffness in joints, Imperial College
London, 2010

Junior PhD students:

•Li, Yanan, Ph.D student, National University of Singapore, from 2008
I helped him to start research on neural network control (Li, Yang, Ge and Lee, 2010, in press) and
then I helped him to study on robotic impedance control using iterative learning approach (Ge, Li and
Yang, 2010, submitted).

•Dai, Shilu, Ph.D student, Exchange student from Northeastern University of China at National
University of Singapore, from 2007
I helped him on the topics of discrete-time adaptive control (Ge, Yang, Dai and Lee, 2009) and (Dai,
Yang, Ge and Lee, 2010, submitted).

To contribute to the excellence of the School and the
I am fully committed
• To serve the needs of each individuals in their
  pursuit of truth,
• To give full support to all passions for excellency
  in science and engineering,
• To help junior faculty members to strive and shine.
Soul with a purpose

       While though control may not
         take any physical forms,
        it exists in every system, and
      serves as the soul of the resided
     system for a meaningful purpose.
• Control bridges different disciplines,
• Control is wanted, and
• Control wants to serve!
 Best wishes
Wish Oxford continue to lead in
1. scientific discovery,
2. economic contribution, and
3. humanity
Oxford has been a dream place for
many bright minds and talented
in the flattened world
I am very honored and privileged to be
invited for the meeting.
Predictive adaptive control

  Bringing Human Neuromotor Intelligence to Robots
Unknow control direction

Bimanutrack at Fraunhofer Institute (left), Tremor
attenuation robotic exoskeleton at IAI-CSIC in Spain (middle)
DLR 7-DOF light weight robot (right)
Want to know more?

Email       X
Telephone   +44 (0)207 594 XXX

To top