[tahoe-dev] GSoC Idea: Locality based dynamic P2P algorithm for distributed introduction
M O Faruque Sarker
writefaruq at gmail.com
Wed Mar 24 12:58:22 PDT 2010
Hello all,
I'm glad to introduce a new project idea from my robotic research project.
Let me introduce myself first. I'm Faruque and doing my PhD on
multi-robot system at Univ. of Wales, Newport, UK. I have some
experiences in writing python code with dbus/multiprocessing modules.
All my dirty codes for robot controllers/multi-robot experiments are
backed up at http://github.com/roboshepherd/
The problem of single point of failure is also common in multi-robot
systems. In our lab, we designed a biology-inspired algorithm for
distributed information spreading and tested with 16 e-puck robots.
This worked better than a centralized counterpart. This algorithm does
not depend on the size of the robots (nodes). It also does not need
to know about peers a priori (no static subscription). It only needs
a common convention/protocol to address the peers which can be put in
the program library.
There are three main parts of this algorithm from a node's point of view:
1. Sensing the local peers ( in case of computer networks this can be
done by following any simple scanning strategy)
2. Listening to own channel to receive incoming messages from peers
(and process that messages)
3. Sending (processed) messages to local peer(s)
All actions of nodes are confined to a certain communication radius or
locality (hence it minimizes computation/communication load). Nodes
can decide to interact with only next peer or a number of peers based
on a pre-set/dynamic logical distances or locality. Unlike robots,
nodes are static. But they can be up and down at any time. So each
node will dynamically decide its peers and continue decentralized
introduction. I used DBus signalling to realize this algorithm. A
conference paper is under review. Please contact me if you like to get
a copy of it.
Please let me know if this can be a GSoC project with your suggestions.
Thanks in advance and best regards,
Faruque
--
¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬
Md Omar Faruque Sarker
PhD Student,
Robotic Intelligence Lab,
Department of Computing,
University of Wales, Newport,
Allt-yr-yn Campus, Newport, NP20 5XR, UK
http://ril.newport.ac.uk/sarker/index.php
Phone:+00441633 432548
¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬
More information about the tahoe-dev
mailing list