Algorithm 1: Dynamic Communication Efficiency Optimization

Input: Set of received information blocks $B$

Set of battery consumptions $BA{T}_{con}^i$

Estimated costs ${\mathrm{Q}}_{\mathrm{consump}}$ and ${\mathrm{Q}}_{\mathrm{obten}}$ (hardware dependent)

Circular buffer $BUFF$ with $N_{hist}$ last $G_{PAT-max}$ evaluations

Circular buffer $ENT$ with $N_{hist}$ last $\overline{H\left(B\right)}$ evaluations

Output: Endpoint configuration

Calculate $Q_{info}$ for the set of received information blocks $B$, $Q_{info}=C_I\left[B\right]$

Calculate $Q_{link}$ using the set of battery consumptions $BA{T}_{con}^i$, $Q_{link}=C_E\left[BA{T}_{con}^i\right]$

Calculate the mean entropy of the information blocks, $\overline{H\left(B\right)}$ Obtain the mean communication efficiency $\overline{\eta }=\eta \left(Q_{info},Q_{link}, {\mathrm{Q}}_{\mathrm{obten}},{\mathrm{Q}}_{\mathrm{consump}}\right)$

if$\overline{\eta }< {\eta }_{threshold}$ or endpoint is not configured then

$CON{F}_{selected}=ALWAYS-ON$

for every server pattern $PAT\in \left\{predefined, Bernoulli, Poisson\right\}$ do

for every cost $Q_i\in \left\{Q_{info}, Q_{link}, Q_{obten}, Q_{consump}\right\}$ do

Calculate $p_i= p\left( Q_i | server=PAT \right)$ using the proposed mathematical model

Obtain $p_{PAT, i}= p\left(server=PAT | Q_i\right)$ using Bayes theorem

end for

Calculate $G_{PAT-total} ={\displaystyle {\displaystyle \prod }_i}G\left(p_{PAT, i}\right)$

end for

Calculate $G_{PAT-max}=max \left\{G_{PAT-total}, being G_{PAT-total}> G_{threshold}\right\}$

Insert $G_{PAT-max}$ in $BUFF$

if $G_{predefined-max}$ is contained in $BUFF$ at least $N_{eval}$ times then

if server allows dynamic scheduling then

$CON{F}_{selected}=DYNAMIC$

else

$CON{F}_{selected}=EXPONENTIAL$

end if

else if $G_{Bernoulli-max}$ is contained in $BUFF$ at least $N_{eval}$ times then

Estimate $p_{ber}$ using the set of received information blocks $B$

if $p_{ber}<0,5$ then

$CON{F}_{selected}=FIXED$

end if

else

if $G_{Poisson-max}$ is contained in $BUFF$ at least $N_{eval}$ times then

$CON{F}_{selected}=EXPONENTIAL$

end if

end if

Insert $\overline{H\left(B\right)}$ in $ENT$

if server allows negotiation then

if $\overline{H\left(B\right)} \in \left[0,\frac{k}{2}\right]$ in $ENT$ at least $N_{eval}$ times then

$COMPRESSION=RLE$

else if $\overline{H\left(B\right)} \in \left[\frac{k}{2},3\frac{k}{2}\right]$ in $ENT$ at least $N_{eval}$ times then

$COMPRESSION=HUFFMAN$

else

if $\overline{H\left(B\right)} \in \left[3\frac{k}{2},k\right]$ in $ENT$ at least $N_{eval}$ times then

$COMPRESSION=RAW$

end if

end if

end if

end if