|
zePPeLIN: Distributed path planning using an overhead camera network
|
|
|
|
Abstract
|
We introduce zePPeLIN, a distributed system designed to address the
challenges of path planning in large, cluttered, and dynamic
environments. The objective is to define the sequence of
instructions to move a ground object from an initial to a final
configuration in the environment. zePPeLIN is based on a set of
wirelessly networked devices, each equipped with a camera,
deployed in environment. Cameras are placed at the ceiling. While
each camera only covers a limited environment portion, the camera
set fully covers the environment through the union of the field
of views. By local message exchanging, the cameras cooperatively
compute the path for the object, which gets moving
instructions from each camera when it enters camera's field of
view. Path planning is performed in a fully distributed way,
based on potential diffusion over local Voronoi skeletons. The
task is made challenging by intrinsic errors in the overlapping
in cameras' field of views. We study the performance of the
system vs. these errors, as well as its scalability for size and
density of the camera network. We also propose a few
heuristics to improve performance and computational and communication
efficiency. We report about extensive simulation experiments and
validation using real devices.
|
|
|
|
Videos
The navigation in the environment.
Once the path is defined, the
system starts the navigation phase. The navigation control process cs
connected to the camera s that has the moving robot in its field of view
starts the navigation phase. Using the camera, process cs tracks the current
position of the two e-pucks and sends them via Bluetooth two independent messages
with the information of the relative movement to be performed, that can be
either a rotation in place or a translation. Each e-puck, when it completes the
required movement, sends a notification message back to cs. Iteratively,
using the current position of the two robots and the next configuration to be
reached, as specified by the locally planned path, the control process
calculates and sends the relative movement that each robot has to perform
next. In this way the camera-robot system operates in a closed loop, such that
it is able to correct possible path implementation errors.
The videos are shown three times faster than reality.
In the videos, when a camera passes the navigation control to the
next camera, a connection path is calculated, which is drawn in yellow.
A camera icon shows which camera is taking control.
Every time the control is passed from one camera to the next one the two robots
rotate in place before starting to move. This is due to the following reason.
In order to make the tracking process simple and scalable, the same color patch
is put on the top of the two e-pucks, such that there is no need to create new
custom patches if new robots are added to the multi-robot structure. However,
this way of proceeding has the drawback that the tracking system has no way to
to distinguish between the two e-pucks, which is needed because each e-puck has
to perform different movements and the control process has to send different
messages to each of them.
Therefore, before starting to send instructions, the control process sends, in
sequence, a message to each one of the known addresses of the two e-pucks,
asking to perform a rotation in place. This allows the control process to
associate each robot to its wireless address.
From the sample videos, it is possible to appreciate that the
robots implementing the instructions received from the overhead camera are able
to follow the calculated path with good precision. The system is able to
effectively correct local actuation errors through the continuous tracking of
the positions and recalculation of the next relative movements to communicate.
Video 1
Download the video.
Video 2
Download the video.
Video 3
Download the video.
|
|
Contact
Webpages:
Address:
| IRIDIA |
|
IRIDIA - ULB
50 Avenue F. Roosevelt - CP 194/9
1050 Bruxelles
Belgium
|
|
| IDSIA |
|
IDSIA
Galleria 2
6928 Manno-Lugano
Switzerland
|
|
|