Simulator Channel Models

How are channel models set?

The channel model is the code inside of the simulator that determines which packets can reach their intended destinations, which are out of range, which are dropped due to noise, and so forth.

The channel model is set in the simulator configuration file, at the line that looks something like:

sim-component = "sim/sim_mote -m circle";

In this case, "circle" is the channel model. It can be any of the following models, which are described in more detail below:

circle -- simple, deterministic 8 meter circle model, POT has no effect
norm_noise -- analytical variable pot, normal noise
uni_noise -- analytical variable pot, uniform noise
emp_avg -- empirical data, average case, indoors
emp_state -- empirical data, stateful, indoors

The Circle Channel Model

The circle model is trivial: any nodes less than 8 meters apart can exchange packets 100% of the time. Nodes further apart can never exchange packets. This is a simple, deterministic model that does not reflect reality at all, but can be useful in getting a first cut application working.

Basic Non-Deterministic Noise Models

The norm_noise and uni_noise models are slightly more realistic. They take into account the value of the POT (potentiometer) that applications might set, and produce a gradual decay in the probability of successful packet delivery as distance increases. The two variants are:

norm_noise -- variable potentiometer with random noise from a normal probability distribution.
uni_noise -- variable potentiometer with random noise from a uniform probability distribution.

Empirical Average Indoors and Empirical Stateful Indoors

These two channel models are based on connectivity measurements performed by Alberto Cerpa with the ipaq/motenic testbed in the LECS lab (indoors, office style). They are a much better improvement to the analytical-only models available thus far, and can help your simulations run under more realistic connectivity conditions.

Empirical Average Indoors (emp_avg) -- is based on average connectivity (empirically obtained) with added noise proportional to the standard deviation (also empirically obtained). The model works for both static and mobile nodes.

Empirical Stateful Indoors (emp_state) -- similar to the above, but it maintains additional state to recreate even more realistic conditions. The model works for static nodes (it requires some modifications to the implementation to make it work with mobile nodes).

Among the "features" these new models provide are:

Variable potentiometer support -- The connectivity values are in direct relation with the potentiometer value set by the hardware user (mote).
Asymmetrical links -- The connectivity of node a to node b (a->b) might be different than from node b to node a (b->a).
Non-isotropical connectivity -- The connectivity is not necessary the same in all the directions (same distance) from the source.
Non-monotonical distance decay -- Nodes that are geographically far away from the source may get better connectivity than nodes that are geographically closer.

These are some of the characteristics encountered when performing real connectivity experiments.

Last modified by jelson, 12 March 2003