Loading the player ...
- Offer Profile
- Our main activity is research
that also drives other ones. We acquire knowledge and find solutions for
open problems. We believe our activities should be guided by the needs of
our society. Therefore we study solutions to problems that would enhance
various aspects of our live and environment.
We have ties with other research groups and we work on international
projects. We are looking forward to establish new connections for
perspective common projects.
- We designed a special 3 degrees of freedom robot
dedicated for skiing using the carving technique. The robot has a complex
sensory and control system that assures stabilization on the ski slope
during skiing and tracks a path marked with race gates and avoids obstacles
The robot is controlled with a hierarchically built multi processor computer
system. The upper level controller is used for navigation, vision processing
and decision making. The upper level controller communicates with a GPS
receiver and an USB based camera. The low level controller deals with skier
stability, joint control and receives information from the sensory system
comprised of an electronic gyroscope, force sensors mounted between the skis
and robot legs and the motor position sensors. Both computers communicate
through Ethernet using UDP protocol. The hierarchical control law was
implemented using the Matlab xpcTarget toolbox with a control sampling
frequency of 1kHz. Vision processing was accomplished using a single camera
with a view angle of 78 degrees. With an image size of 320 x 200 pixels we
were able to obtain the object recognition and localization at 30 fps. The
distance between the current robot position and the next gate was calculated
using the camera orientation and the estimated size of the gate mark. We
used blue and red gate fences, respectively. The color defined whether the
robot-skier should approach them from the left or from the right side
depending on the color. The desired turn radii were estimated from the
localization information, and the angular acceleration. The velocity in the
sagittal plane was obtained by the GPS sensor with an accuracy of 0.1 m/s
and sample time of 0.25 s.
We performed several experimental runs on the ski slope. Our experiments
showed that the robot was able to follow the desired path marked by any
combination of four race gates providing that the distance between the gates
allowed the robot to execute the turn.
Skiing Robot on stage