Information Technology Reference
In-Depth Information
a centralized server for planning the inspection routes. Likewise, a stable and
reliable communications system is needed.
One of the handicaps in this type of underground environments is the di-
culty to keep continuous communication among the members that conform the
team. This requires (1) high degree of autonomy for each member of the group
and (2) all the members must share all the possible information when the envi-
ronment allows the communication between themselves. Despite new techniques
are growing [1,5], empirical evaluations [17] shown that, for example, the soil
composite or oil present are a decisive aspect.
In this paper, we introduce a redundant, robust and fault-tolerant method,
which is able to perform inspection tasks on environments described above. In-
spection tasks assume knowledge about the environment. We model this knowl-
edge like a connected graph to represent the infrastructure of the sewer or pipe
network. The agents of the team compete against each other through auctions
to prevail the best travel routes and then complete the main task.
The paper is structured as follow. First, we describe how the proposed method
works and the requirements that must fulfilled. Second, we expose the different
experiments in a brief comparative with other methods. Finally, we comment the
conclusions and possible improvements in our method as well as some proposals
for future works.
2 Description of the Proposed Method
2.1 Objective
The aim of the proposed method is to inspect, in a finite time, a pipe or sewer
environment modeled as a graph by the collaboration of a team of agents. We
consider the inspection task as the action to visit each of the edge that comprise
the graph by some of the team's members of the multi-robot team. To accomplish
this, agents can make use of communications for coordination with the rest of the
team when the circumstances allow it. The method proposed is able to overcome
failures in the communication or in the robots' hardware. In the worst case, a
single agent should be able to, eventually, visit all edges of the graph.
2.2 Prerequisites
Modeling a pipe network or a sewerage system like a graph allows us to work with
powerful data structures and make use of graph theory. Thus, nodes represent
intersections and edges represent corridors or tunnels.
When an agent reaches a node, it knows which node is. If an agent wants to
go from the node A to the node B, eventually it will arrive at node B directly
from node A, i.e. an agent placed in the node A knows which edge has to be
selected to travel to the node B.
One characteristic of our method is that each agent could start to work at
any moment, it is not needed that the agents start altogether. Furthermore, it
is possible to lose agents during the inspection task.
Search WWH ::
Custom Search