| 85 | | %\subsubsection{Configuration} |
| 86 | | % |
| 87 | | %For comparative analysis, we consider two related works: |
| 88 | | %Vigorito \cite{vigorito07} and Hsu \cite{hsu06} . Before we run each |
| 89 | | %algorithm, we configured each algorithm so that it produced |
| 90 | | %the best possible results. |
| | 86 | \subsection{Comparison} |
| | 87 | |
| | 88 | For comparative analysis, we consider two related works: |
| | 89 | Vigorito \cite{vigorito07} and Hsu \cite{hsu06} . Before we run each |
| | 90 | algorithm, we configured each algorithm so that it produced |
| | 91 | the best possible results. |
| | 92 | |
| | 93 | |
| 105 | | %%Thus, we set the input parameters as follows: |
| 106 | | %%$\beta$ = 0.5 and $\alpha$ = 0.01. |
| 107 | | %%\begin{itemize} |
| 108 | | %%\item $\beta$ : At $\beta = 0.5$, $|\overline{E_{tra}}|$ |
| 109 | | %%is minimized to 0.028. |
| 110 | | %%\item $\alpha$ : At $\alpha = 0.01$, $|\overline{E_{tra}}|$ |
| 111 | | %%is minimized to 0.510 while keeping the zerodays to 0. |
| 112 | | %%\end{itemize} |
| | 108 | %Thus, we set the input parameters as follows: |
| | 109 | %$\beta$ = 0.5 and $\alpha$ = 0.01. |
| | 110 | %\begin{itemize} |
| | 111 | %\item $\alpha$ : exponential moving average factor |
| | 112 | %for the historical average of duty-cycle ($\overline{\mu}$). |
| | 113 | %$\alpha$ is set to 0.01, where $|\overline{E_{tra}}|$ |
| | 114 | %is minimized to 0.510 and zerodays is kept to 0. |
| | 115 | %\item $\beta$ : exponential moving average factor |
| | 116 | %between the duty-cycle calculation ($\mu$) and |
| | 117 | %its historical average ($\overline{\mu}$). |
| | 118 | %$\beta$ is set to 0.5, where $|\overline{E_{tra}}|$ |
| | 119 | %is minimized to 0.028. |
| | 120 | %\item $\alpha$ : At $\alpha = 0.01$, $|\overline{E_{tra}}|$ |
| | 121 | %is minimized to 0.510 while keeping the zerodays to 0. |
| | 122 | %\end{itemize} |
| 118 | | %%Figure~\ref{fig:hsu_algorithm} shows that the |
| 119 | | %%performance of Hsu's algorithm depends on the exponential |
| 120 | | %%moving average factor $\alpha$ that is used to predict |
| 121 | | %%the daily harvested solar energy $\overline{E_s}$ |
| 122 | | %%from the historical measurement $E_s$. We set the $\alpha$ |
| 123 | | %%so that the algorithm minimizes the difference from the |
| 124 | | %%target energy level, which is at $\alpha = 0.75$. |
| 125 | | % |
| | 128 | %Figure~\ref{fig:hsu_algorithm} shows that the |
| | 129 | %performance of Hsu's algorithm depends on the exponential |
| | 130 | %moving average factor $\alpha$ that is used to predict |
| | 131 | %the daily harvested solar energy $\overline{E_s}$ |
| | 132 | %from the historical measurement $E_s$. We set the $\alpha$ |
| | 133 | %so that the algorithm minimizes the difference from the |
| | 134 | %target energy level, which is at $\alpha = 0.75$. |
| | 135 | |
| | 136 | %%% THIS IS THE LAST GRAPH |
| 127 | | %%\begin{tabular}{|p{0.45\textwidth}|} |
| 128 | | %%\hline |
| 129 | | %%\begin{scriptsize} |
| 130 | | %%\begin{algorithmic}[1] |
| 131 | | %%%\STATE $\theta$ $\leftarrow$ $[2,-1,1]$ |
| 132 | | %%%\STATE $B$ $\leftarrow$ $B_{init} / B_{max}$ |
| 133 | | %%%\STATE $B^*$ $\leftarrow$ $B_{target} / B_{max}$ |
| 134 | | %%%\STATE $\rho$, $\mu$, $\overline{\mu}$ $\leftarrow$ $DC_{init}$ |
| 135 | | %%%%\STATE $\mu$ $\leftarrow$ $DC_{init}$ |
| 136 | | %%%%\STATE $\overline{\mu}$ $\leftarrow$ $DC_{init}$ |
| 137 | | %%%\STATE $\phi$ $\leftarrow$ $[B, \mu, -B^*]$ |
| 138 | | %%\LOOP |
| 139 | | %% \STATE $B$ $\leftarrow$ $Bat / B_{max}$ |
| 140 | | %% \STATE $\theta$ $\leftarrow$ $\theta + \frac{\mu}{\phi \cdot \phi} \times (B - \phi \cdot \theta) \times \phi$ |
| 141 | | %% \STATE $\mu$ $\leftarrow$ $\min(1, \max(0, \frac{B^* - \theta(1) \times B + \theta(3) \times B^*}{\theta(2)}))$ |
| 142 | | %% \STATE $\phi$ $\leftarrow$ $[B, \mu, -B^*]$ |
| 143 | | %% \STATE $\overline{\mu}$ $\leftarrow$ $\overline{\mu} + \alpha \times (\mu - \overline{\mu})$ |
| 144 | | %% \STATE $\rho$ $\leftarrow$ $\beta \times \mu + (1 - \beta) \times \overline{\mu}$ |
| 145 | | %% \STATE $1/T_{samp}$ $\leftarrow$ $(1-\rho)/T_{sampmax} + \rho/T_{sampmin}$ |
| 146 | | %% \STATE $1/T_{report}$ $\leftarrow$ $(1-\rho)/T_{reportmax} + \rho/T_{reportmin}$ |
| 147 | | %%\ENDLOOP |
| 148 | | %%\end{algorithmic} |
| 149 | | %%\end{scriptsize} |
| 150 | | %%\\ |
| 151 | | %%\hline |
| 152 | | %%\end{tabular} |
| | 138 | %\begin{tabular}{|p{0.45\textwidth}|} |
| | 139 | %\hline |
| | 140 | %\begin{scriptsize} |
| | 141 | %\begin{algorithmic}[1] |
| | 142 | %%\STATE $\theta$ $\leftarrow$ $[2,-1,1]$ |
| | 143 | %%\STATE $B$ $\leftarrow$ $B_{init} / B_{max}$ |
| | 144 | %%\STATE $B^*$ $\leftarrow$ $B_{target} / B_{max}$ |
| | 145 | %%\STATE $\rho$, $\mu$, $\overline{\mu}$ $\leftarrow$ $DC_{init}$ |
| | 146 | %%%\STATE $\mu$ $\leftarrow$ $DC_{init}$ |
| | 147 | %%%\STATE $\overline{\mu}$ $\leftarrow$ $DC_{init}$ |
| | 148 | %%\STATE $\phi$ $\leftarrow$ $[B, \mu, -B^*]$ |
| | 149 | %\LOOP |
| | 150 | % \STATE $B$ $\leftarrow$ $Bat / B_{max}$ |
| | 151 | % \STATE $\theta$ $\leftarrow$ $\theta + \frac{\mu}{\phi \cdot \phi} \times (B - \phi \cdot \theta) \times \phi$ |
| | 152 | % \STATE $\mu$ $\leftarrow$ $\min(1, \max(0, \frac{B^* - \theta(1) \times B + \theta(3) \times B^*}{\theta(2)}))$ |
| | 153 | % \STATE $\phi$ $\leftarrow$ $[B, \mu, -B^*]$ |
| | 154 | % \STATE $\overline{\mu}$ $\leftarrow$ $\overline{\mu} + \alpha \times (\mu - \overline{\mu})$ |
| | 155 | % \STATE $\rho$ $\leftarrow$ $\beta \times \mu + (1 - \beta) \times \overline{\mu}$ |
| | 156 | % \STATE $1/T_{samp}$ $\leftarrow$ $(1-\rho)/T_{sampmax} + \rho/T_{sampmin}$ |
| | 157 | % \STATE $1/T_{report}$ $\leftarrow$ $(1-\rho)/T_{reportmax} + \rho/T_{reportmin}$ |
| | 158 | %\ENDLOOP |
| | 159 | %\end{algorithmic} |
| | 160 | %\end{scriptsize} |
| | 161 | %\\ |
| | 162 | %\hline |
| | 163 | %\end{tabular} |
| | 164 | |
| | 165 | |
| 188 | | For comparative analysis, we consider two related works |
| 189 | | on energy-harvesting-aware duty-cycling algorithms: |
| 190 | | Vigorito \cite{vigorito07} and Hsu \cite{hsu06} . |
| 191 | | To compare the performance of these algorithms with ours, |
| 192 | | we consider the following metrics: |
| | 216 | |
| | 217 | |
| | 218 | %\subsection{Simulation Results} |
| | 219 | |
| | 220 | To compare the performance of Vigorito's algorithm |
| | 221 | with ours, we consider the following metrics: |
