AHDI - An ad hoc and a distributed infrastructure for controlling a group of moving agents

In this proposal we consider the development of a general distributed infrastructure with
two components: an ad-hoc communication layer and a decision making component.
This infrastructure can be used by a cooperative group of moving agents to
complete a given global task while communicating using the unstable ad-hoc media.
The global task is assume to include the minimization of a given performance measure.
We will test the usefulness of the proposed infrastructure
considering tasks such as scan, sample, guard or intercept a non-cooperative group of targets.
Such an infrastructure can be tested by simulating a set of of unattended ground vehicles (UGVs)
attempting to autonomously complete one of the above tasks.

The dynamic changes in the communication environment require
distributed algorithms, without relying on a central agent, which may
constitute a single point-of-failure and is exposed to communication breaks.

Addressing objective ICT-2007.3.7.b

This project aims to develop a general distributed infrastructure to be used
by a group of moving agents attempting to complete a given global task.
The infrastructure contains two components: an ad hoc communication layer
and a basic distributed algorithm. Both components are specially designed
to support simple and easy implementation of the above global task.
The global task is modeled as a task assignment (TA)
algorithm that controls the movements of the agents overcoming frequent
communicating breakdown caused by the use of the ad-hoc networks.
The tasks considered in this proposal include: scanning and locating
desired targets, patrolling (e.g., patrolling a set of oil pipelines),
frequent sampling of moving targets, rescue,
forest fire detection/fighting and interception.
In particular, we assume that the agents are unattended ground vehicles (UGVs),
dispersed in an unknown theater and are aimed at engaging multiple moving
targets.

The first layer of the proposed project addresses the need to develop
a reliable and scalable communication algorithm over ad-hoc networks.
The unique dynamic nature of ad-hoc communication between a group of
moving agents (nodes) attempting to complete a given task calls for a special novel
type of ad-hoc routing algorithm. This include the following:
1- The algorithm is required under all conditions to support a
   reasonable connectivity and satisfactory level of quality of service between nodes.
2- The proposed algorithm will try to equally disperse communication sessions
   in the 2D-plain in order to overcome mutual blockings of calls all trying to
   ``cross'' through a relatively small ``segment'' of nodes. This aspect has not
   been considered in previous works.
3- The proposed algorithm should support the implementation of distributed TA
   algorithms by maintaining a sufficiently stable structure (such as a minimal
   set of spanning trees) over the set of nodes.
   Whenever possible, the spanning trees will merge into a single tree that
   clusters the maximum number of nodes.


The second layer of the proposed project is the distributed
TA algorithms that will be based on the dynamic spanning trees structure that is maintained
by the ad-hoc communication layer.
All TA algorithms will use distributed relaxation methods to complete a given
global task.
Each node will combine inputs from its parent and children nodes,
optimizing its own solution separately from other ``remote'' or possibly
currently disconnected parts of the network.
A global and more optimized solution will eventually obtained due to the
relaxation processes at the roots of the underlying spanning trees.
This will result in a true distributed control of the moving agents allowing them
to coordinate their movements in spite of the dynamic changes in their
communication topology.

A suitable simulator will be developed to test the proposed  TA algorithm in real-world scenarios.
The simulator will support 3D ground surface including reflections of radio
waves from obstacles. We will not assume GPS abilities and relay on the ad-hoc
communication to generate approximate coordinates.
The simulator will be a common software combining all the separate
projects done under this proposal.

The project does not consider the actual driving of the UGVs and will use a
simplified model of obstacles. The main focus is on the communication layer and
the distributed control.

Ad hoc, agents, distributed tasks, optimizations.

Experties are either in ad-hoc communication, wireless networks, distributed algorithms and agents systems.

0- communication layer to create  the dynamic tree topology connecting the moving agents.
1- general distributed  relaxation component.
2- distributed algorithm for each task : scan, sample, gard or intercept  using the dynamic tree.
3- simulation environment.
4- lab environment to test the algorithms.
5- User interfaces how to program and control the system.

Academic researchers and commercial communication companies.

Yes