Simple questions for sensor network.

• How the sensor knows its position? 

In current project, each sensor node have a unique identifier, which has been assigned by the base station (gate way) during the initial processing. The project set up a 4*5 matrix landscape to deploy 20 sensor notes. Every sensor nodes occupy a fixed one of the twenty places, i.e., each note has a fix coordinates. 

To learn they neighbours, all nodes are required to broadcast their own positions information when the system start. After this step, each sensor node has a blueprint of the whole network’s position. 

• How the sensor knows the target’s position? 

In previous work, each sensor node is equipped by an infrared ray. When a sensor node finds a target by infrared ray method, it will measure the distance (d1) via sonar. Next step, the current sensor node notices one of its neighbours to measure the distance (d2). Under the triangulation method, with two nodes position (x1, y1), (x2, y2) and d1, d2, we can get two possible points for the target. Considering on the target should always pass the network’s boundary from outside into inside, we can omit the point outside the deployment area. When the target starts to move, this problem becomes a tracking issue.

Note that in this work, this method assume the target is found by the boundary nodes firstly. But if we use the third sensor node, a trilateration method is useful to determine the target’s position, i.e. use three known points and three known edges to determine a node in a 2D- plane. 

All of above are proposed for 2-dimension. But here is a question that, in a concrete application, the sensors may be deployed in a 3-dimension space. In this case, to evaluate the target’s position, we should use some sensing data with direction property. 

• How the sensors track the target currently? 

In current project, we only consider single target tracking. To detect the target in all directions, a sensor node is equipped with six sonar sensor, which been ranged in an array with a 60 degree angle. To avoid intervening, we assume that there is only one single sensor node is working in each time step. 

We use EKF to estimate the target’s route. As we know, EKF is a recursive filter, what we need is to provide some initial conditions.
The whole process can be describe as follow, when a target enter into the sensor network area, one of the boundary sensor nodes (N1) will do the first job. Use the localization method above, N1 will get the target’s position. Then it will use EKF to estimate the target’s next step, and pass all messages to one of its neighbour. The neighbour node repeats this process to hand the messages to next one.