For 2011 Science festival, Toulouse town hall funded us to realize an artistic performance with
two well-known artists from Toulouse together with the humanoid robot HRP-2.
The artists:
This video demonstrates a new fast re-planning method applied to the humanoid robot HRP-2.
The method is described in more details in the following papers:
In this video, technologies developped in the frame of a french national project
called R-Blink
are used in a European Project called
ROBOT@CWE.
This project aims at performing collaborative tasks between a robot and a human. In this particular case,
the robot is holding a panel with a human. It gently follows the human leading the way to the table where
the panel should be stored.
Movie of the final demonstrator (50 Mo).
Click here to watch the movie of the first prototype (47 Mo).
Direct human-humanoid robot interaction
In this movie, HRP-2 is gently following a human through the information given by its force-sensor located
in the wrist. The hands are controlled using a compliant controller with its apparent mass
equivalent to the one of the robot. In the movie you can see different modes of walking:
straight, rotation and lateral motions but there no
special treatement between each of them. Self-collision is avoided thanks to an approximation
function filtering the
feasible and non-feasible foot-steps. You can also see the robot performing whole-body
motion when it uses its chest-joint
to follow the hand of the user and stay balanced.
Click here to watch the movie (45 Mo).
Strategies for Humanoid Robots to Dynamically Walk over Large Obstacles
In this video you will see how HRP-2 can cross over while walking obstacles as high
as 15 cm. This video shows in addition dynamical simulation demonstrating
that the robot can mechanically cross over
25 cm obstacles.
Click here to watch the movie (16 Mo).
Online Object Search with a Humanoid Robot
In this video a real HRP-2 is exploring autonomously a room. It builds a map of the environment
using its vision stereoscopic system and find the best pose to acquire the maximum of information.
When it succeeded to plan a feasible solution the robot acquire new information.
In simulation you can see that this algorithm performs very well to find an object.
Click here to watch the movie (5 Mo).
Integrating Walking and Vision to increase Humanoid Autonomy
In this video, we shows several key technologies developped to increase a robot autonomy:
dynamical walk to cross over obstacles, foot-steps planning based on vision information,
Visual SLAM using walking pattern information and real-time whole body motion generation
based on vision.
Click here to watch the movie (28 Mo).