\section{Introduction}

\begin{sloppypar}
The rapid proliferation of smartphones creates both challenges and new
opportunities for wireless networks. On one hand, smartphones compete for the
same limited spectrum already crowded with other devices. On the other hand,
because smartphones are \textit{always on} but \textit{mostly idle}, they are
ideal for observing the network conditions on behalf of nearby active wireless
devices. When used for continuous network adaptation,
offloading measurements to inactive clients avoids
disrupting active sessions, a capability that has not been adequately
exploited by other systems using client-side feedback. When used for network
monitoring and debugging, smartphones provide more valuable measurements than
planned site surveys, since the data that
smartphones provide is continuous and representative of wireless conditions
experienced by users while surveys are neither. We refer to these approaches
collectively as \textbf{c}rowdsourcing \textbf{a}ccess \textbf{n}etwork
\textbf{s}pectrum \textbf{a}llocation using \textbf{s}martphones, or
\textbf{CANSAS}.
\end{sloppypar}

\begin{sloppypar}
We are currently developing a prototype system called \PS{} that implements CANSAS for \wifi{}
networks. It collects measurements from passive smartphones to
improve network performance. \PS{} also captures a variety of
measurements generated naturally by smartphones as they discover and connect to
networks---valuable data that is currently discarded.
\end{sloppypar}