Event Based Routing in Sensor Networks
Weilin/Zhendong/Gus
Abstract
In this proposal, we present a novel routing protocol for sensor networks, which does not
require location service, or a unique ID for individual node.
Problem
Unlike the Internet, sensor networks are tightly coupled to the physical environment. In
addition, the task of the sensor network is usually highly related to its physical
environment. However, current routing protocols could not deal with the changes in the
environment directly. Therefore, to build a sensor network which can interface with its
environment is very efficient.
Solution
Our approach is an event-based routing protocol for sensor networks. In the context,
event can be described as a certain change of the environment, such as temperature,
noise, color, and lightness increase/decrease to a threshold. The format of event is:
(EventID, Evaporation model, diffusion model).
 Applications define events for nodes, and nodes just repeat the events. A certain
object can be located by the combination of several events.

In this model, each node has no unique identification. Each node just transfers
data and control information by broadcasting to its immediate neighbors.

No location service or knowledge is required.

Routing traffic is indexed by event, thus a higher scalability can be achieved.
The whole process concludes several steps:
Event Pheromone Diffuse Procedure
In the event source node, when a certain event is triggered, it will send an event to its
neighbors in the following format:
(EventId, density, TTL, hop-count)
Events are cumulatively diffused in the sensor network, nodes update the event density
according to the evaporation model. And update happens when necessary.
The node where the event has been triggered broadcasts event pheromones to its
neighbors, which records events into its routing table, like following:
EventID
Density
Evaporation Arrivial
TTL
Hop –
Diffused
model
time
count
model
1
100
.5%
10
20
2
-10
2
200
.1%
20
10
3
*2%
When its neighbors receives event information, it delays a certain amount of time, then
broadcasts to its own neighbors. The time of delay is defined:
Delay = x * hop_count,
This is to make the event diffusion in reverse proportion to the distance to the source
node.
The new event message will be:
(EventID, density’, TTL’, hop_count’)
In which TTL’ = TTL – Deley, density’ = f(density) => diffusion model.
Routing: from user to source
A
First, the node will broadcast with its own event density.
Event_density_inquery(eventID, current density)
All the neighbors will reply with its event density. A will pick up a node with the highest
density, then broadcasting with this value, then broadcasts again.
Event_highest_density(eventId, highest_density, expected_density, data)
Nodes matching exactly the same density will actually accept this message, then forward
this message in the same way, until the event density reaches a certain amount (the
destination).
Evaluation
Since the node is broadcasting whenever there is an event proactively, the destination is
tightly coupled to the event, instead of the node itself. For example, when event
E1(intruder detected) happens near node A, but the intruder may not be at the same
location when the system decides to response. In this case, event based routing will guide
the correct control information to B (where the intruder is), instead of A, according to the
evaporation and pheromone model in this routing protocol.
Future works
Proactive broadcasting maybe is not good for energy conservation.
Sensors are pre-programmed to execute a certain kind of task by the attributes, not
flexible.
How to define attributes – this is a concern because definition of attributes can affect the
diffusion of data greatly.