I. Introduction
COGNITIVE spectrum sharing was recently studied to accommodate growing demands for wireless broadband access, which can alleviate the problem of under-utilization of licensed spectrum. Spectrum sharing techniques can be generally classified into three categories: interweave, underlay, and overlay [1]. For the interweave spectrum sharing, the secondary system can opportunistically access spectrum holes. For the spectrum underlay, secondary users (SUs) transmit simultaneously with primary users (PUs) under the constraint that interference caused by the SUs on the PUs must be below a certain threshold. For the spectrum overlay, SUs actively help primary data transmission in exchange for a spectrum access in time domain [2], spatial domain [3], or frequency domain [4]. The locations of SUs are usually fixed or restricted into a small area without suffering interference from other concurrent transmissions. It is nontrivial to extend the cooperative spectrum sharing to the secondary ad-hoc networks, as the topology changes frequently and the interference suffers from uncertainties caused by both random user locations and channel fadings.