|
Course code:
wb2433-03
|
|
Course name:
Humanoid Robots
|
|
This concerns a Course
|
|
ECTS credit points:
3
|
|
Faculty of Mechanical Engineering and Marine Technology
|
|
Section of Man-Machine Systems
|
|
Lecturer(s):
M.
Wisse, R.Q. van der Linde, P. Jonker, P. van Lith, M. Verhaegen
|
Tel.:
015 - 27
86585
|
|
Catalog data:
Humanoid Robots, Robotics, Vision systems, Walking robots, Haptics, Robot Soccer
|
Course year:
|
MSc 1st year
|
|
Semester:
|
2A
|
|
Hours per week:
|
4
|
|
Other hours:
|
|
|
Assessment:
|
Oral exam
|
|
Assessment period:
|
2A
|
|
(see academic
calendar)
|
|
|
Prerequisites (course codes):
BSc. requirements
|
|
Follow up (course codes):
|
|
Detailed description of topics:
Humanoid robots are the research topic of the
future, and partially already today. This course is organized around the
central problem in humanoid robot design; they must operate fully
autonomously. This results in design constraints such as energy efficiency
and autonomous control. The course will treat the following topics:
|
|
Course material:
-
Reader (not yet available)
|
|
References from literature:
|
|
Remarks assessment, entry requirements, etc.:
The students are required to have a personal
interest and motivation for robotics.
|
|
Learning goals:
The student must be able to:
-
design a modular system architecture
for autonomous robots. For each of the software or hardware modules, the
student can describe (1) the function of the module, (2) the services
that the module provides to higher-ranking modules, (3) the services
that the module requires from lower-ranking modules, (4) the type(s) of
interface(s) that the module requires
-
describe which functions a (any)
multibody dynamics simulation package fullfills, which types of
algorithms are used in the package, and which typical problems can arise
(accuracy, instability) and where these problems originate. Also, the
student can describe the similarities between PD controllers and
mechanical spring-damper systems
-
describe the various existing methods
to control two-legged walking robots. The student knows and is able to
calculate the two most common performance criteria, namely stability
(plus robustness) and efficiency. The student can describe by which
means the robustness can be increased
-
explain the principle of reinforcement
learning and the special case of Q-learning. The student is able to set
up a learning controller (i.e. defining the length and conditions of
learning trials, the inputs and outputs, and the reward structure). The
student can describe the effects of various reward settings and explore
rates, and name potential pittfalls and advantages
-
select electric DC motors and gear
boxes for a given required torque-velocity pattern, and accounting for
motor inertia effects and gear energy losses. The student can list the
type of sensors required to measure the full state of a robot system.
The student can explain why it is difficult to measure the absolute
orientation of the system and provide a solution. The student can also
explain how one can create a “series-elastic actuation” system
-
apply an image processing library to
perform low-level image processing algorithms and higher-level feature
detection, enabling the automated detection of for example the location
and size of an orange ball in an image. The student can explain why a
color space other than RGB is used, and how the feature data can be used
to obtain 3D information about the object of interest
-
describe how images of faces can be
processed in order to obtain information about the face expression
|
|
Computer use:
|
|
Laboratory project(s):
|
|
Design content:
|
|
Percentage of design:
25%
|