Thomas Kunz
 Systems and Computer Engineering
11
Prediction Algorithm Simulation
lImplemented in NS2 to validate the algorithm
uNo modification to the routing protocol yet
uCompare the predicted and “real link breakage time”
lParameter setting
uCritical Time: the time needed to rebuild a new route
lResult:
uNodes with active link have above 90% prediction accuracy
uNode movement pattern changes may cause false prediction
We implemented the prediction algorithm is ns2 to validate. We did not modify the routing protocol yet at this step. When the node calculated a link breakage time, it record in the trace file. Then we can compare it with the real packet dropped time which is always later than link breakage  time. If they match, we say the algorithm prediction is correct.

The critical time is a parameter that the time needed for source node to receive the prediction Route Error message and build a new route. It is different for different network size and routes with different hops. Here we use a static value.

We ran the simulation for different mobility scenarios. The result shows nodes with active link have above 90% prediction accuracy. The active link is the link that has unicast communication over it in one second. So the prediction algorithm may have enough packets to calculate.
Also, if nodes change their movement patterns during this time, for example, they are moving closer, then it will cause false prediction. It only account for less than 5%.