Cognitive radio network is a concept of wireless communication for mobile devices that offers the possibility to exploit the unused spectrum resources opportunistically. These networks bring out the need for new solutions that mitigate the spectrum management issue. However, existing works do not focus on devices mobility whereas serious problems arise when users are mobile specifically about their provided quality of services. In this work, we study spectrum sharing and spectrum handoff for mobile secondary users (SUs) and we propose a novel approach that can be executed by a mobile SU when traveling through wireless networks. The proposed solution is inspired from multi-agent system auctions and integrates a learning module which accelerates SUs’ spectrum bands allocation. One of the main contributions of this paper is the realistic implementation of the learning based auction and the interesting results obtained through a network discrete event simulator. Results prove that our proposal enhances spectrum utilization and guarantees users satisfaction.